Long-term and short-term customer behavior prediction method and device

Abstract

The present invention provides a method and device for predicting long-term and short-term customer behavior, which can be used in the field of artificial intelligence technology. The method includes: determining whether there is a first transaction data of a user within a first preset time period before the current time based on a preset trigger condition; if so, performing a short-term behavior prediction on the transaction data through a preset short-term prediction model to obtain a short-term behavior prediction result; if not, obtaining the second transaction data of the user within a second preset time period before the current time, performing a long-term behavior prediction on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result, the first preset time period is less than the second preset time period, and the present invention can improve the accuracy of customer behavior prediction. Another object of the present invention is to provide a long-term and short-term customer behavior prediction device.

Description

Technical Field

[0001] The present invention relates to the field of behavior prediction technology, in particular to the field of artificial intelligence technology, and more particularly to a method and device for predicting long-term and short-term customer behavior. Background Art

[0002] In financial institutions such as banks, historical data of various dimensions of customers can be used to generate customer portraits to predict customer transaction behavior. The main functions of predicting customer transaction behavior can be divided into two aspects: one is to predict customers' potential business needs and transaction behaviors, and assist customers to handle business quickly; the other is to analyze customers' product preferences and assist customer service personnel or bank systems in product marketing.

[0003] However, with the continuous development of banking business, customer touchpoints continue to increase, and customer data is growing explosively. The existing customer transaction behavior prediction methods have certain limitations. The current customer transaction behavior prediction methods only consider the customer's long-term static data, and only make behavior and preference predictions based on the customer's entire historical data. They ignore the analytical bias that may still be caused by short-term changes in customer behavior, resulting in the customer portrait model being unable to accurately capture and predict customer dynamics. Summary of the invention

[0004] One object of the present invention is to provide a method for predicting long-term and short-term customer behavior to improve the accuracy of customer behavior prediction. Another object of the present invention is to provide a device for predicting long-term and short-term customer behavior. Another object of the present invention is to provide a computer device. Another object of the present invention is to provide a readable medium.

[0005] In order to achieve the above objectives, the present invention discloses a method for predicting long-term and short-term customer behavior, comprising:

[0006] Determine whether there is first transaction data of the user within a first preset time period before the current time based on a preset trigger condition;

[0007] If yes, a short-term behavior prediction is performed on the transaction data using a preset short-term prediction model to obtain a short-term behavior prediction result;

[0008] If not, obtain the second transaction data of the user within a second preset time period before the current time, and perform long-term behavior prediction on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result, and the first preset time period is smaller than the second preset time period.

[0009] Preferably, after obtaining the short-term behavior prediction result, the method further comprises:

[0010] Acquire second transaction data of the user within a second preset time period, and perform long-term behavior prediction on the second transaction data using a preset long-term prediction model to obtain a long-term behavior prediction result;

[0011] A comprehensive prediction result is obtained based on the short-term behavior prediction result and the long-term behavior prediction result.

[0012] Preferably, obtaining a comprehensive prediction result based on the short-term behavior prediction result and the long-term behavior prediction result specifically includes:

[0013] Determine a first push result among the short-term behavior prediction results;

[0014] Determining a second push result in the long-term behavior prediction result;

[0015] A comprehensive prediction result is formed according to the first push result and the second push result.

[0016] Preferably, the method further comprises the step of preforming the long-term prediction model:

[0017] Determine the long-term basic data and long-term behavioral data in the user's historical data;

[0018] The long-term prediction model is obtained by training a preset machine learning model based on the long-term basic data and the long-term behavioral data based on a binary classification algorithm.

[0019] Preferably, further comprising:

[0020] forming independent variable feature importance according to the long-term basic data;

[0021] Collecting customer data corresponding to the importance of the independent variable features;

[0022] The stability of the long-term prediction model is calculated according to the customer data, and whether the long-term prediction model needs to be updated is determined according to the stability. If so, the long-term prediction model is retrained.

[0023] Preferably, the method further comprises the step of preforming the short-term prediction model:

[0024] Determine the short-term basic data and short-term behavior data in the user's historical data;

[0025] The short-term prediction model is obtained according to the short-term basic data and the short-term behavior data through collaborative filtering algorithm and expert rules.

[0026] Preferably, further comprising:

[0027] Monitor users’ short-term trading behavior;

[0028] When there is short-term trading behavior, collect the user's short-term trading data;

[0029] The short-term prediction model is retrained according to the short-term transaction data.

[0030] The present invention also discloses a long-term and short-term customer behavior prediction device, comprising:

[0031] A trigger determination module, used to determine whether there is first transaction data of the user within a first preset time period before the current time based on a preset trigger condition;

[0032] A first prediction module, for, if yes, performing short-term behavior prediction on the transaction data by using a preset short-term prediction model to obtain a short-term behavior prediction result;

[0033] If no, the second prediction module obtains the second transaction data of the user within a second preset time period before the current time, and performs long-term behavior prediction on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result, and the first preset time period is smaller than the second preset time period.

[0034] The present invention also discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0035] When the processor executes the program, the method described above is implemented.

[0036] The present invention also discloses a computer readable medium having a computer program stored thereon.

[0037] When the program is executed by a processor, the method described above is implemented.

[0038] The long-term and short-term customer behavior prediction method of the present invention determines whether there is a first transaction data of a user within a first preset time period before the current time based on a preset trigger condition. If so, a short-term behavior prediction is performed on the transaction data through a preset short-term prediction model to obtain a short-term behavior prediction result. If not, the second transaction data of the user within a second preset time period before the current time is obtained, and a long-term behavior prediction is performed on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result. The first preset time period is less than the second preset time period. The present invention determines whether the preset trigger condition is currently met. When the trigger condition is met, it is determined whether there is a first transaction data of the user within the first preset time period before the current time, that is, it is determined whether the user has short-term transaction behavior. If there is short-term transaction behavior, a short-term behavior prediction is performed on the first transaction data to obtain a short-term behavior prediction result. If the user does not have the first transaction data of a short-term transaction within the first preset time period, the second transaction data of the user within the second preset time period before the current time is obtained. The second transaction data is data formed by the user's long-term transaction behavior. The user can perform a long-term behavior prediction on the second transaction data to obtain a long-term behavior prediction result. By predicting the behavior of user transaction data, the user's short-term behavior prediction results or long-term behavior prediction results can be obtained, and the obtained short-term behavior prediction results or long-term behavior prediction results are fed back to the management personnel, so that the management personnel can push corresponding products or services to the user according to the user's behavior prediction results. Therefore, the present invention overcomes the defect that the existing customer behavior prediction only performs calculations and analysis on long-term historical data, and provides a combined analysis method of long-term and short-term transaction data based on current customer behavior changes. On the basis of long-term customer transaction behavior prediction, a comprehensive analysis of short-term customer behavior prediction is additionally added, and dynamic customer behavior prediction is performed with a more variable comprehensive analysis method, replacing the cumulative static customer behavior prediction for all data, thereby increasing the hit rate of the analysis results, improving the accuracy of customer behavior prediction, and improving the prediction accuracy of user preferences to more accurately push relevant financial products and services to users, thereby improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 A structural diagram showing a specific embodiment of the long-term and short-term customer behavior prediction system of the present invention;

[0041] Figure 2A flowchart showing a specific embodiment of the method for predicting long-term and short-term customer behavior of the present invention;

[0042] Figure 3 A flowchart showing a specific embodiment S400 of the method for predicting long-term and short-term customer behavior of the present invention;

[0043] Figure 4 A flowchart showing a specific embodiment S420 of the method for predicting long-term and short-term customer behavior of the present invention;

[0044] Figure 5 A flowchart showing a specific embodiment S010 of the method for predicting long-term and short-term customer behavior of the present invention;

[0045] Figure 6 A flowchart showing a specific embodiment S500 of the method for predicting long-term and short-term customer behaviors of the present invention;

[0046] Figure 7 A flowchart showing a specific embodiment S020 of the method for predicting long-term and short-term customer behavior of the present invention;

[0047] Figure 8 A flowchart showing a specific embodiment S600 of the method for predicting long-term and short-term customer behaviors of the present invention;

[0048] Fig. 9 A structural diagram showing a specific embodiment of the long-term and short-term customer behavior prediction device of the present invention;

[0049] Fig.10 A schematic diagram showing the structure of a computer device suitable for implementing an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] It should be noted that the long-term and short-term customer behavior prediction method and device disclosed in the present application can be used in the field of artificial intelligence technology, and can also be used in any field outside the field of artificial intelligence technology. The application field of the long-term and short-term customer behavior prediction method and device disclosed in the present application is not limited.

[0052] In order to facilitate the understanding of the technical solution provided by the present application, the relevant contents of the technical solution of the present application are first explained below. The current customer transaction behavior prediction method only considers the long-term static data of the customer, and only predicts the behavior and preference based on the customer's entire historical data, ignoring the analysis bias that may still be caused by the customer's short-term behavior changes, resulting in the customer portrait model being unable to accurately capture and predict customer dynamics. Customer transaction behavior and preferences will change over time, and short-term behavioral changes will not have a significant impact on the analysis results of long-term customer portraits due to the small amount of data. At this time, a model that takes into account both long-term and short-term customer data is needed to respond to customer changes in a timely manner. Especially for bank branches, at the moment when online business is developing rapidly, most of the customer's transactions will be completed through online channels. If they come to offline branches to handle business, it is usually more due to their special life changes, behavioral changes, preference changes, etc. in the short term.

[0053] For example, a customer has been purchasing wealth management products at bank branches for ten years. However, due to travel needs, he purchased foreign currency online in the past week and needed to withdraw foreign currency cash at a bank branch. If only historical data is analyzed, the customer's foreign exchange purchase operation in a short period of time will not have a significant change in his customer profile. When the customer arrives at the store, the customer profile generated after system identification will still recommend wealth management products to him, and his possible foreign currency withdrawal needs cannot be accurately obtained, resulting in the analysis results not being able to play a good predictive and auxiliary role, and unable to effectively improve customer experience and service efficiency.

[0054] The long-term and short-term customer behavior prediction method provided by the embodiment of the present invention overcomes the defect that the existing customer behavior prediction only performs calculation and analysis on long-term historical data. Based on the current customer behavior changes, it provides a combined analysis method of long-term and short-term transaction data. On the basis of long-term customer transaction behavior prediction, it also adds a comprehensive analysis of short-term customer behavior prediction, and uses a more variable comprehensive analysis method to perform dynamic customer behavior prediction, replacing the cumulative static customer behavior prediction for all data, thereby increasing the hit rate of the analysis result, improving the accuracy of customer behavior prediction, and improving the prediction accuracy of user preferences so as to more accurately push relevant financial products and services to users, thereby improving user experience.

[0055] Figure 1 is a schematic diagram of the structure of the long-term and short-term customer behavior prediction system provided by an embodiment of the present invention. Figure 1 As shown, the long-term and short-term customer behavior prediction system provided by the embodiment of the present invention includes a transaction system 1 and a long-term and short-term customer behavior prediction device 2.

[0056] The transaction system 1 can provide transaction services for users and form transaction data based on the transaction behavior of users. The transaction data can include information such as transaction time, transaction type, transaction details, and transaction amount. Therefore, it can be determined whether the user has short-term transaction behavior through the transaction time in the transaction data.

[0057] The long-term and short-term customer behavior prediction device 2 is used to determine whether there is a first transaction data of a user within a first preset time period before the current time based on a preset trigger condition; if so, a short-term behavior prediction is performed on the transaction data through a preset short-term prediction model to obtain a short-term behavior prediction result; if not, the second transaction data of the user within a second preset time period before the current time is obtained, and a long-term behavior prediction is performed on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result, and the first preset time period is less than the second preset time period.

[0058] The following uses the long-term and short-term customer behavior prediction device 2 as an example to illustrate the implementation process of the long-term and short-term customer behavior prediction method provided by the embodiment of the present invention. It is understandable that the execution subject of the long-term and short-term customer behavior prediction method provided by the embodiment of the present invention includes but is not limited to the long-term and short-term customer behavior prediction device 2.

[0059] According to one aspect of the present invention, this embodiment discloses a method for predicting long-term and short-term customer behavior. Figure 2 As shown, in this embodiment, the method includes:

[0060] S100: Determine whether there is first transaction data of a user within a first preset time period before the current time based on a preset trigger condition.

[0061] Specifically, the preset trigger condition can be a preset fixed time point, which can be a fixed time every day. For example, the fixed time can be selected from 2:00 or 3:00 a.m. every day or every few days when the production transaction system is relatively idle, or batch processing time such as end-of-day batch, end-of-month batch, and online small batch. Of course, in actual applications, other fixed times can also be selected to start transaction detection to determine whether the user has the first transaction data within a short period of time (first preset time period). The preset trigger condition can also be a specific transaction behavior of the customer, such as the customer swiping the card to get a number at the bank business point, the bank self-service machine inserting the card to log in, or the bank monitoring system capturing and identifying the customer. The preset trigger condition can also be automatically triggered by a specific node, such as a bank account manager generating a customer portrait of the customer who comes to the store through the terminal or generating a special node such as personalized product recommendation at the end of the channel transaction. Of course, the preset trigger condition can also be triggered by manual operation of the management personnel, such as manual triggering when the management personnel inquire about the basic information of the customer or generate a customer portrait. In actual applications, those skilled in the art can set the preset trigger condition according to actual needs, and the present invention is not limited to this.

[0062] S200: If yes, a short-term behavior prediction is performed on the transaction data using a preset short-term prediction model to obtain a short-term behavior prediction result.

[0063] S300: If not, obtain the second transaction data of the user within a second preset time period before the current time, and perform long-term behavior prediction on the second transaction data using a preset long-term prediction model to obtain a long-term behavior prediction result, and the first preset time period is smaller than the second preset time period.

[0064] Specifically, if the customer has no transaction behavior in the short term, that is, there is no first transaction data, then only the long-term prediction model needs to be introduced to predict the user's behavior. If the customer has no transaction data in the short term, it means that there is no behavior change, or the model cannot be analyzed based on the short-term transaction data. In this case, there is no need to introduce a short-term prediction model to predict the customer's behavior, and only the long-term behavior prediction results can be used.

[0065] It should be noted that the first transaction data and the second transaction data may include information such as transaction time, transaction type, transaction details, and transaction amount. Therefore, it is possible to determine whether the user has short-term transaction behavior through the transaction time in the transaction data, that is, whether the first transaction data exists in the customer's transaction data. Among them, the first preset time period can be a time interval such as S days, S weeks, or S months before the current time, where S is a positive integer. The time length of the second preset time period is greater than the time length of the first preset time period. Similarly, the second preset time period can be a time interval such as S days, S weeks, or S months before the current time that is greater than the time length of the first preset time period. Of course, in actual applications, the first preset time period and the second preset time period of other time intervals can also be selected.

[0066] The long-term and short-term customer behavior prediction method of the present invention determines whether there is a first transaction data of a user within a first preset time period before the current time based on a preset trigger condition. If so, a short-term behavior prediction is performed on the transaction data through a preset short-term prediction model to obtain a short-term behavior prediction result. If not, the second transaction data of the user within a second preset time period before the current time is obtained, and a long-term behavior prediction is performed on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result. The first preset time period is less than the second preset time period. The present invention determines whether the preset trigger condition is currently met. When the trigger condition is met, it is determined whether there is a first transaction data of the user within the first preset time period before the current time, that is, it is determined whether the user has short-term transaction behavior. If there is short-term transaction behavior, a short-term behavior prediction is performed on the first transaction data to obtain a short-term behavior prediction result. If the user does not have the first transaction data of a short-term transaction within the first preset time period, the second transaction data of the user within the second preset time period before the current time is obtained. The second transaction data is data formed by the user's long-term transaction behavior. The user can perform a long-term behavior prediction on the second transaction data to obtain a long-term behavior prediction result. By predicting the behavior of user transaction data, the user's short-term behavior prediction results or long-term behavior prediction results can be obtained, and the obtained short-term behavior prediction results or long-term behavior prediction results are fed back to the management personnel, so that the management personnel can push corresponding products or services to the user according to the user's behavior prediction results. Therefore, the present invention overcomes the defect that the existing customer behavior prediction only performs calculations and analysis on long-term historical data, and provides a combined analysis method of long-term and short-term transaction data based on current customer behavior changes. On the basis of long-term customer transaction behavior prediction, a comprehensive analysis of short-term customer behavior prediction is additionally added, and dynamic customer behavior prediction is performed with a more variable comprehensive analysis method, replacing the cumulative static customer behavior prediction for all data, thereby increasing the hit rate of the analysis results, improving the accuracy of customer behavior prediction, and improving the prediction accuracy of user preferences to more accurately push relevant financial products and services to users, thereby improving user experience.

[0067] In a preferred embodiment, Figure 3 As shown, the method further comprises, after obtaining the short-term behavior prediction result, S400:

[0068] S410: Acquire second transaction data of the user within a second preset time period, and perform long-term behavior prediction on the second transaction data using a preset long-term prediction model to obtain a long-term behavior prediction result.

[0069] S420: Obtaining a comprehensive prediction result based on the short-term behavior prediction result and the long-term behavior prediction result.

[0070] It is understandable that, in the case where a customer has short-term first transaction data, a short-term prediction model can be introduced to predict the customer's first transaction data to obtain a short-term behavior prediction result. Furthermore, in this preferred embodiment, in order to further improve the accuracy of customer behavior prediction, a long-term prediction model can be simultaneously introduced to predict the customer's second transaction data to obtain a long-term behavior prediction result. Since the customer has first transaction data in the short term, it means that there is a possibility of behavioral changes in the short term. At the same time, the customer's long-term transaction data also reflects the customer's overall transaction behavior trend on the one hand. Therefore, while introducing short-term behavior prediction, long-term behavior prediction is additionally introduced to conduct a comprehensive analysis of customer behavior to obtain a comprehensive prediction result, which is used to adapt to short-term customer behavior changes and correct the result deviation that may occur when relying solely on long-term customer behavior prediction.

[0071] In a preferred embodiment, Figure 4 As shown, the S420 obtains the comprehensive prediction result according to the short-term behavior prediction result and the long-term behavior prediction result, specifically including:

[0072] S421: Determine the first push result among the short-term behavior prediction results.

[0073] S422: Determine a second push result in the long-term behavior prediction result.

[0074] S423: Form a comprehensive prediction result according to the first push result and the second push result.

[0075] Specifically, in this preferred embodiment, comprehensive prediction results are obtained through long-term and short-term customer behavior predictions, and customers' potential business needs and transaction behaviors are predicted, so that banks and other financial institutions can push customer behavior prediction results to staff through various channels to assist staff in quickly handling business for customers. At the same time, customers' product preferences can also be analyzed through prediction results, so that banks and other financial institutions can push them to staff or marketing terminals through various channels to assist product marketing.

[0076] In a specific example, forming a comprehensive prediction result based on the first push result and the second push result can be based on actual usage needs, defining a specific value N as the first push result of the short-term behavior prediction result, and adding the short-term behavior prediction results of the short-term customer portrait to the final comprehensive prediction result according to the number of N, then the number of short-term behavior prediction results in the comprehensive prediction result is N, if the total push result is M, then the remaining push results of the M push results except the N short-term behavior prediction results are the second push results of the long-term behavior prediction result. In other examples, according to actual usage needs, any value R between 0 and 1 can be defined as the push ratio coefficient of the short-term behavior prediction result, and the short-term behavior prediction results of the short-term customer portrait can be added to the final comprehensive prediction result according to the proportion of R, then the proportion of short-term behavior prediction results in the final comprehensive prediction result is R, and the proportion of long-term behavior prediction results is 1-R. More preferably, it is recommended that the proportion of long-term and short-term prediction results in the comprehensive prediction result is 3:2, which can better encompass changes in customer behavior and increase prediction accuracy. For example, if the bank forms a trigger condition when the customer arrives at the store and takes a number, and pushes 5 services that may need to be handled to the customer based on the comprehensive prediction result formed by the first push result and the second push result, the long-term behavior prediction result and the short-term behavior prediction result can be pushed in a ratio of 3:2. Among them, the short-term behavior prediction results and the long-term behavior prediction results include the services or products that the customer is likely to handle. First, according to the customer's long-term asset structure and historical preferences, push 3 financial products that may be needed, and second, according to the customer's short-term foreign exchange purchase record, push the settlement and cash withdrawal services of specific foreign currencies.

[0077] In a preferred embodiment, Figure 5 As shown, the method further includes a step S010 of pre-forming the long-term prediction model:

[0078] S011: Determine the long-term basic data and long-term behavior data in the user's historical data.

[0079] S012: According to the long-term basic data and the long-term behavioral data, a preset machine learning model is trained based on a binary classification algorithm to obtain the long-term prediction model.

[0080] Specifically, the data acquisition cycle and modeling observation point of the long-term prediction model can be determined. According to the actual use needs of banks and other financial institutions, a certain point in time L in the past can be defined as a long-term time node, and the point L to date can be defined as the second preset time period. The cycle for the model to obtain the second transaction data can be L days, L weeks, L months, etc. from the current date; it can also be the recent L days, L weeks, L months, etc. of the current date. At the same time, according to the actual use needs of banks and other financial institutions, a certain point in time T from point L to the present is defined in a personalized way in the model as a modeling observation point. In the long-term range, the period before point T is the observation period, and the period after point T is the performance period.

[0081] Furthermore, basic customer data during the observation period is obtained as the model independent variable. During the observation period within the defined long-term range, basic data such as customer basic information, purchased product information, risk information, association information, behavior information, preference information, and evaluation information are obtained as independent variables of the long-term prediction model.

[0082] The customer behavior data during the performance period is used as the model dependent variable. During the performance period within the defined long-term range, the customer's banking interaction, business processing, product purchase and other behavioral data are counted as the dependent variable of the long-term prediction model.

[0083] The long-term prediction model in this preferred embodiment uses a binary classification algorithm (observing whether a customer has purchased a certain product, 1 for purchase, 0 for non-purchase, and training a binary classification model based on customer profile data and purchase behavior data) as an auxiliary algorithm for modeling. Based on the basic data and behavior data of customers in a long-term range, the binary classification algorithm is used as an auxiliary to generate a relationship model between the behavior data and basic data of a specific customer in a long-term range to obtain a long-term prediction model, that is, the specific laws by which each basic data constitutes the customer behavior are used as the basis for future prediction and continuous correction of customer behavior.

[0084] In a preferred embodiment, Figure 6 As shown, the method further includes S500:

[0085] S510: Forming independent variable feature importance according to the long-term basic data.

[0086] S520: Collect customer data corresponding to the importance of the independent variable feature.

[0087] S530: Calculate the stability of the long-term prediction model according to the customer data, determine whether the long-term prediction model needs to be updated according to the stability, and if so, retrain the long-term prediction model.

[0088] Specifically, in order to control and improve the prediction accuracy of the model, the feature importance of the independent variable, i.e. the basic data, can be generated on the basis of training the long-term prediction model, which is used as the basis for determining whether the long-term prediction model needs to be updated. Among them, the binary classification algorithm is mainly divided into a tree model and a linear model. The tree model can directly output the feature importance after training, and the linear model can use the regression coefficient as the feature importance.

[0089] According to the generated feature importance, collect customer data of important features. Sort the generated feature importance from large to small, and the features that are ranked in the top 10%-35% and whose importance generated in the previous step is significantly not 0 are important features. After continuously collecting customer data of important features, use it as the data input formula for calculating the stability of the model. Then, calculate the PSI (Population Stability Index) population stability index based on the collected important feature data to determine the stability of the model features. The PSI index calculation formula is PSI = sum((actual proportion - expected proportion) × ln(actual proportion / expected proportion)), that is, PSI = sum[(Ac-Ex) × ln(Ac / Ex)]. Use the PSI index to determine the stability of the model and determine whether the model needs to be updated.

[0090] Among them, according to the actual usage needs of banks and other financial institutions, a specific PSI value U is personalized and defined in the model as the model feature stability threshold. Once the PSI value is greater than U, it means that the characteristics of the model have changed significantly, and it is necessary to re-acquire data to train the long-term prediction model; when the PSI value is less than or equal to U, it means that the characteristics of the model have not changed significantly, and the current model can continue to be used for prediction.

[0091] In a preferred embodiment, Figure 7 As shown, the method further includes a step S020 of pre-forming the short-term prediction model:

[0092] S021: Determine short-term basic data and short-term behavior data in the user's historical data.

[0093] S022: Obtain the short-term prediction model according to the short-term basic data and the short-term behavior data through collaborative filtering algorithm and expert rules.

[0094] Specifically, the basic data and behavior data of customers can be obtained within a defined short-term range. Since the data in a short-term range is usually less, it is difficult to accurately capture the laws in the training machine learning model, and the prediction results are not of reference value. Therefore, there is no need to set modeling observation points. Instead, collaborative filtering algorithms and expert rules are used to establish a short-term customer portrait model.

[0095] The short-term customer portrait model of the present invention adopts a collaborative filtering algorithm (calculating the similarity between the customer to be predicted and the known customer, and recommending to the customer similar products that the known customer has purchased) as an auxiliary algorithm for modeling. Specific rules are introduced as auxiliary rules for establishing a short-term customer portrait model. The short-term customer portrait model of the present invention adopts expert rules (making rules based on expert experience for prediction) as auxiliary rules for modeling. Based on the basic data and behavioral data of the customer within a short-term range, a collaborative filtering algorithm and expert rules are used as an aid to generate a short-term prediction model for a specific customer within a short-term range.

[0096] In a preferred embodiment, Figure 8 As shown, the method further includes S600:

[0097] S610: Monitor the user's short-term trading behavior.

[0098] S620: When there is short-term trading behavior, collect the user's short-term trading data.

[0099] S630: Retrain the short-term prediction model according to the short-term transaction data.

[0100] It is understandable that in order to control and improve the prediction accuracy of the model, customer data can be continuously collected to improve the accuracy of the model prediction. Specifically, customer data within a short period of time can be continuously collected to improve the accuracy of the collaborative filtering algorithm and continuously update the short-term customer portrait model. The update cycle can be triggered by the touch points of the customer's interaction with the bank system (such as swiping a card to get a number at the store, inserting a card to log in to the self-service machine, being captured and recognized by the camera, etc.); it can also be executed regularly by the bank system (such as end-of-day batches, end-of-month batches, online small batches, etc.); it can also be automatically triggered by specific nodes of the bank system (such as the customer service manager's handheld device generates a customer portrait at the store, and the channel transaction ends to generate personalized product recommendations, etc.); it can also be triggered by manual operations of bank back-end staff (such as querying basic customer information, generating specific customer portraits, etc.).

[0101] The present invention overcomes the defect that the existing customer behavior prediction only performs calculation and analysis on long-term historical data. Based on the current customer behavior changes, it provides a combined analysis method of long-term and short-term transaction data. On the basis of long-term customer transaction behavior prediction, it also adds a comprehensive analysis of short-term customer behavior prediction, which can make timely responses to customer behavior changes in the short term, and use a more changeable comprehensive analysis method to perform dynamic customer behavior prediction, replacing the cumulative static customer behavior prediction for all data, thereby increasing the hit rate of the analysis results, improving the accuracy of customer behavior prediction, and improving the accuracy of user preference prediction, so as to more accurately push relevant financial products and service efficiency to users, thereby improving product marketing success rate and user experience.

[0102] Based on the same principle, this embodiment also discloses a long-term and short-term customer behavior prediction device. Fig. 9 As shown, in this embodiment, the device includes a trigger determination module 11, a first prediction module 12 and a second prediction module 13.

[0103] The trigger determination module 11 is used to determine whether there is first transaction data of the user within a first preset time period before the current time based on a preset trigger condition.

[0104] The first prediction module 12 is used to perform short-term behavior prediction on the transaction data by using a preset short-term prediction model to obtain a short-term behavior prediction result.

[0105] The second prediction module 13 is used to obtain the second transaction data of the user within a second preset time period before the current time, and perform long-term behavior prediction on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result, and the first preset time period is smaller than the second preset time period.

[0106] In a preferred embodiment, the first prediction module 12 is further used to obtain second transaction data of the user within a second preset time period after obtaining the short-term behavior prediction result, perform long-term behavior prediction on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result, and obtain a comprehensive prediction result based on the short-term behavior prediction result and the long-term behavior prediction result.

[0107] In a preferred embodiment, the first prediction module 12 is specifically used to determine a first push result in the short-term behavior prediction result; determine a second push result in the long-term behavior prediction result; and form a comprehensive prediction result based on the first push result and the second push result.

[0108] In a preferred embodiment, the second prediction module 13 is further used to pre-form the long-term prediction model: determine the long-term basic data and long-term behavior data in the user's historical data; and train a preset machine learning model based on a binary classification algorithm according to the long-term basic data and the long-term behavior data to obtain the long-term prediction model.

[0109] In a preferred embodiment, the second prediction module 13 is further used to form the independent variable feature importance based on the long-term basic data; collect customer data corresponding to the independent variable feature importance; calculate the stability of the long-term prediction model based on the customer data, and determine whether the long-term prediction model needs to be updated based on the stability, and if so, retrain the long-term prediction model.

[0110] In a preferred embodiment, the first prediction module 12 is further used to pre-form the short-term prediction model: determine the short-term basic data and short-term behavior data in the user's historical data; and obtain the short-term prediction model based on the short-term basic data and the short-term behavior data through a collaborative filtering algorithm and expert rules.

[0111] In a preferred embodiment, the first prediction module 12 is further used to monitor the short-term trading behavior of the user; when there is short-term trading behavior, collect the short-term trading data of the user; and retrain the short-term prediction model according to the short-term trading data.

[0112] Since the principle of solving the problem by the device is similar to that of the above method, the implementation of the device can refer to the implementation of the method and will not be repeated here.

[0113] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer device, and specifically, the computer device may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0114] In a typical example, the computer device specifically includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method executed by the client as described above is implemented; or, when the processor executes the program, the method executed by the server as described above is implemented.

[0115] Reference below Fig.10 , which shows a schematic diagram of the structure of a computer device 600 suitable for implementing an embodiment of the present application.

[0116] like Fig.10 As shown, the computer device 600 includes a central processing unit (CPU) 601, which can perform various appropriate operations and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 to a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the system 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0117] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal feedback device (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 606 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed as needed as the storage section 608.

[0118] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 609, and / or installed from the removable medium 611.

[0119] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0120] For the convenience of description, the above device is described in terms of functions and is described separately in various units. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0121] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0122] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0123] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0124] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0125] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0126] The present application may 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 may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0127] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0128] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for predicting long-term and short-term customer behavior, characterized in that: include: Determine whether there is first transaction data of the user within a first preset time period before the current time based on a preset trigger condition; wherein the preset trigger condition includes: a preset fixed time point, a specific transaction behavior of the customer, a specific node, and an operation behavior of a manager; the first transaction data includes: transaction time, transaction type, transaction details, and transaction amount; If yes, a short-term behavior prediction is performed on the transaction data using a preset short-term prediction model to obtain a short-term behavior prediction result; If not, obtain the second transaction data of the user within a second preset time period before the current time, and perform long-term behavior prediction on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result. The first preset time period is smaller than the second preset time period, and the second transaction data includes: transaction time, transaction type, transaction details, and transaction amount.

2. The method for predicting long-term and short-term customer behavior according to claim 1, characterized in that: Further comprising after obtaining the short-term behavior prediction result: Acquire second transaction data of the user within a second preset time period, and perform long-term behavior prediction on the second transaction data using a preset long-term prediction model to obtain a long-term behavior prediction result; A comprehensive prediction result is obtained based on the short-term behavior prediction result and the long-term behavior prediction result.

3. The method for predicting long-term and short-term customer behavior according to claim 2, characterized in that: The obtaining of the comprehensive prediction result according to the short-term behavior prediction result and the long-term behavior prediction result specifically includes: Determine a first push result among the short-term behavior prediction results; Determining a second push result in the long-term behavior prediction result; A comprehensive prediction result is formed according to the first push result and the second push result.

4. The method for predicting long-term and short-term customer behavior according to claim 1, characterized in that: The method further comprises the step of preforming the long-term prediction model: Determine the long-term basic data and long-term behavioral data in the user's historical data; The long-term prediction model is obtained by training a preset machine learning model based on the long-term basic data and the long-term behavioral data based on a binary classification algorithm.

5. The method for predicting long-term and short-term customer behavior according to claim 4, characterized in that: Further including: forming independent variable feature importance according to the long-term basic data; Collecting customer data corresponding to the importance of the independent variable features; The stability of the long-term prediction model is calculated according to the customer data, and whether the long-term prediction model needs to be updated is determined according to the stability. If so, the long-term prediction model is retrained.

6. The method for predicting long-term and short-term customer behavior according to claim 1, characterized in that: Further comprising the step of preforming the short-term prediction model: Determine the short-term basic data and short-term behavior data in the user's historical data; The short-term prediction model is obtained according to the short-term basic data and the short-term behavior data through collaborative filtering algorithm and expert rules.

7. The method for predicting long-term and short-term customer behavior according to claim 6, characterized in that: Further including: Monitor users’ short-term trading behavior; When there is short-term trading behavior, collect the user's short-term trading data; The short-term prediction model is retrained according to the short-term transaction data.

8. A long-term and short-term customer behavior prediction device, characterized in that: include: A trigger determination module is used to determine whether there is a first transaction data of a user within a first preset time period before the current time based on a preset trigger condition; wherein the preset trigger condition includes: a preset fixed time point, a specific transaction behavior of a customer, a specific node, and an operation behavior of a manager; the first transaction data includes: transaction time, transaction type, transaction details, and transaction amount; A first prediction module, for, if yes, performing short-term behavior prediction on the transaction data by using a preset short-term prediction model to obtain a short-term behavior prediction result; The second prediction module is used to obtain the second transaction data of the user within a second preset time period before the current time, and perform long-term behavior prediction on the second transaction data through a preset long-term prediction model to obtain a long-term behavior prediction result. The first preset time period is less than the second preset time period. The second transaction data includes: transaction time, transaction type, transaction details, and transaction amount.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

10. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images