CN115631006A - Method and device for intelligently recommending bank products, storage medium and computer equipment - Google Patents
Method and device for intelligently recommending bank products, storage medium and computer equipment Download PDFInfo
- Publication number
- CN115631006A CN115631006A CN202211371022.XA CN202211371022A CN115631006A CN 115631006 A CN115631006 A CN 115631006A CN 202211371022 A CN202211371022 A CN 202211371022A CN 115631006 A CN115631006 A CN 115631006A
- Authority
- CN
- China
- Prior art keywords
- user
- target
- purchase intention
- data
- bank product
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
- G06N5/041—Abduction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/02—Banking, e.g. interest calculation or account maintenance
Abstract
The application relates to a method, a device, a storage medium and computer equipment for intelligently recommending bank products, wherein the method comprises the following steps: acquiring user data of a target bank product in a preset historical time range, wherein the user data comprises static basic data related to each user and transaction flow data formed by each user in a transaction process; classifying the user data to generate a multi-dimensional user label; constructing a user image of at least one service scene according to the user label; predicting the purchase intention value of each user for the target bank product in the target time range by combining the user portrait based on a Bayesian model; and recommending the target bank product to the target user according to the purchase intention value. According to the method and the device, the purchasing intention value of the user is predicted by utilizing the Bayesian model and the user portrait, the safety in the predicting process of the purchasing intention value can be enhanced, the predicting precision of bank products is further improved, the purchasing experience of the target user is improved, and meanwhile, the labor cost is saved.
Description
Technical Field
The application relates to the field of artificial intelligence, in particular to a method, a device, a storage medium and computer equipment for intelligently recommending bank products.
Background
With the emergence of new technologies such as big data and artificial intelligence, new insurance products and services are generated continuously, and the decision flow and thinking mode of users for the new insurance products and services are more and more intelligent. Compared with the traditional product driving mode of banks, the bank insurance expansion business in the new environment is bound to develop towards customization and intellectualization, which is both opportunity and challenge for banks. For example, although experience shows that the conversion cost of the quasi-user is far lower than the cost of acquiring a new user, the bank has less information about the quasi-user and has a limited analysis means, so that effective conversion cannot be performed on the user.
In the process of rapid development of the internet and deep application of big data technology, a large amount of mass data is accumulated in various industries, and at the moment, subjective analysis and prediction of purchasing willingness of a user are difficult to carry out. Therefore, the precision of marketing opportunities can be effectively improved by utilizing a big data technology in the related technology, and a mode that big data analysis is matched with manual targeted service is formed. Specifically, in the related art, interference of subjective factors such as artificial emotions can be avoided by constructing a deep learning prediction model. The deep learning model has strong data processing capacity and information mining capacity, adopts a computer to carry out automatic analysis, and has quick response.
However, although the big data technology can help the bank to subdivide and insights existing users, accurately know key requirements of the existing users and establish a prediction model, a conventional deep learning prediction model established based on the big data technology generally directly learns mapping between original data and purchasing will to be predicted through a deep neural network, the learning process is a black box and lacks interpretability, and the prediction result has unreliable risk, so that the conversion of the bank to the users is possibly failed, and user experience is influenced.
Disclosure of Invention
In view of this, the present application provides a method for intelligently recommending a bank product and a related device thereof, which can enhance the security in the prediction process of a purchase intention value, further improve the accuracy of bank product prediction, improve the purchase experience of a target user, and save the labor cost at the same time.
According to an aspect of the application, a method for intelligently recommending bank products is provided, and the method for intelligently recommending the bank products comprises the following steps: acquiring user data of a target bank product in a preset historical time range, wherein the user data comprises static basic data related to each user and transaction flow data formed by each user in a transaction process; classifying the user data to generate multi-dimensional user tags respectively corresponding to the users; constructing user images of at least one service scene corresponding to each user respectively according to the user tags; predicting a purchase intention value of each user for the target bank product in a target time range by combining the user portrait based on a Bayesian model; and recommending the target bank product to the target user according to the purchase intention value.
According to another aspect of the present application, there is provided a bank product intelligent recommendation device, including: the data acquisition module is used for acquiring user data of a target bank product within a preset historical time range, wherein the user data comprises static basic data related to each user and transaction flow data formed by each user in a transaction process; the label generation module is used for classifying the user data and generating multi-dimensional user labels respectively corresponding to the users; the image construction module is used for constructing user images of at least one service scene respectively corresponding to the users according to the user tags; the prediction module is used for predicting the purchase intention value of each user for the target bank product by combining the user portrait based on a Bayesian model; and the recommending module is used for recommending the target bank product to the target user according to the purchase intention value.
In some embodiments of the present application, there may also be provided a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method as described above when executing the computer program.
In some embodiments of the present application, there may also be provided a computer-readable storage medium having stored thereon a computer program which, when run on a computer, causes the computer to perform the steps of the method as described above.
The user data of the target bank product in a preset historical time range is classified, a multi-dimensional user label is generated, then a user portrait of at least one corresponding business scene is constructed, then the purchase intention value of each user to the target bank product in the target time range is predicted by combining the user portrait based on a Bayesian model, and finally the target bank product is recommended to the target user according to the purchase intention value.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
Fig. 1 is a schematic application environment diagram illustrating a method for intelligently recommending bank products according to an embodiment of the present application.
Fig. 2 is a flowchart illustrating a method for intelligently recommending bank products according to an embodiment of the present application.
Fig. 3 shows a schematic diagram of a bayesian model of an embodiment of the present application.
Fig. 4 shows a schematic diagram of a bayesian neural network of an embodiment of the present application.
Fig. 5 shows a schematic diagram of an intelligent bank product recommendation device according to an embodiment of the present application.
Fig. 6 shows a schematic structural diagram of a computer device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. It should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The method for intelligently recommending the bank products can be applied to the application environment shown in fig. 1. Among other things, the terminal 101 may communicate with the server 102 over a network. Specifically, the server 102 may obtain user data in the database, bank staff may operate through the terminal 101, send the user data to the server 102 through the terminal 101, the server 102 performs classification processing on the user data in response to the user data, generates multi-dimensional user tags corresponding to the users, constructs a user representation of at least one service scene corresponding to each user according to the user tags, predicts a purchase intention value of each user for the target bank product based on a bayesian model and the user representation, and finally recommends the target bank product to the target user according to the purchase intention value.
The terminal 101 may be a personal computer, a notebook computer, a smart phone, a tablet computer, or other terminal devices, or may also be a portable wearable device, a smart television, or other intelligent terminals.
The server 102 may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing basic cloud computing services such as a cloud service, a cloud database, cloud computing, a cloud function, cloud storage, a Network service, cloud communication, a middleware service, a domain name service, a security service, a Content Delivery Network (CDN), a big data and artificial intelligence platform, and the like.
The following are detailed below. It should be noted that the following description of the embodiments is not intended to limit the preferred order of the embodiments.
Specifically, as shown in fig. 2, a schematic flow chart of a method for intelligently recommending bank products according to an embodiment of the present application is shown, and the method is described by taking an example of applying the method to the server in fig. 1, where the method may include:
step S1: acquiring user data of a target bank product in a preset historical time range, wherein the user data comprises static basic data related to each user and transaction flow data formed by each user in a transaction process;
wherein the user data may include static base data associated with each user. The static basic data may include basic information of the user, such as name, age, occupation, location, and identification number. The transaction flow data may include information such as transaction time, transaction object, transaction amount, transaction appendix, transaction details, etc. For example, the transaction item may be a deposit transaction, and the corresponding transaction object is a bank product related to the deposit transaction. In the transaction flow data, the bank products can be further subdivided, and information such as product names, product numbers, application ranges and the like related to the bank products can be obtained in advance.
It should be noted that the transaction flow data may be a part of dynamic behavior data in the user data. The dynamic behavior data may also include information such as browsing data of the user when using an application developed by the bank itself. The historical time range may be preset as needed, and the application is not limited.
Step S2: classifying the user data to generate multi-dimensional user tags respectively corresponding to the users;
specifically, the user data may be classified based on Natural Language Understanding (NLU) and text clustering, so as to generate multi-dimensional user tags respectively corresponding to the users. It can be understood that there are many implementations of natural language processing and text clustering, and the application is not limited to how the natural language processing and text clustering are implemented.
Further, classifying the user data to generate multidimensional user tags respectively corresponding to the users may include:
step S21: preprocessing the user data to obtain preprocessed user data;
specifically, the preprocessing may include verification of integrity of the data set, filling of missing values in the data set, normalization of feature values in the data set, filtering of useless data, and removal of products and users with no behavioral interaction. In practical applications, the user data may have some data attributes missing and indeterminate, so that the integrity of the user data may be verified before. In case the user data is incomplete, missing values in the data set may be filled, and the filling values may be selected as required. Then, deduplication processing can be carried out, useless data can be filtered out, and meanwhile products and users with no behavior interaction can be removed. In order to solve the inconsistency of different types of data, each feature data in the data set can be further normalized, and the normalized user data is used for further processing. It is understood that each specific step of the pretreatment can be set according to actual needs, and the application is not limited to the order of the pretreatment.
Step S22: and performing classification processing according to the preprocessed user data to generate multi-dimensional user tags respectively corresponding to the users, wherein the dimensions of the user tags at least comprise tag themes, tag sources and tag attributes.
In the application, the user tags may include multiple dimensions, and the user tag division criteria of different dimensions may be set as needed. It should be understood by those skilled in the art that the dimension of the user tag is not limited to the existing description, and the dimension of the user tag can be adjusted as needed.
In one example, the user tags may be sorted by tag topic dimension. For example, intuitive and descriptive ideas can be used to decompose the user attributes to obtain subject information reflected by the tag content, and further induce the subject of the user tag. The topics of the generalized user tags may include user status, product possession, user interaction, user management, and the like.
Further, the user tags may be sorted by tag source dimension. For example, different types of user tags may be sourced from different sources, some user tags may be sourced from within the bank, and some user tags may be sourced from data outside the bank. User tags of different sources are also distinguished in actual use. User tags originating from within the bank may be used directly in the description, processing or reprocessing process of the tag content, while user tags originating from outside the bank may need to be further processed for reuse in the description, processing or reprocessing process of the tag content. User tags may include objective fact class tags, statistical analysis class tags, and evaluation prediction class tags based on whether tags from different sources can be used directly for description of tag content or in the processing and reprocessing processes. The objective fact type label can be a label directly generated according to the existing information, the statistical analysis type label can be a label generated after the existing label is subjected to statistical induction, and the evaluation prediction type label can be a user label predicted at a future moment on the basis of the existing label.
Further, the user tags may also be classified according to tag attribute dimensions. Specifically, the user tags may be classified into static tags and dynamic tags according to whether the attributes of the user tags change over time. The static label may include information that does not change over time such as user industry type, user size, user registered capital, etc., and the dynamic label may include information that frequently changes over time such as channel preferences for purchasing bank products, quantity characteristics of held products, etc.
By dividing the user labels from different data dimensions, the user portrait can be further constructed based on structured analysis, meanwhile, the interference of useless data on the analysis process is avoided, the accuracy of the user portrait is improved, the construction efficiency is improved, and therefore effective support is provided for the final realization of banking business requirements.
And step S3: constructing user images of at least one service scene corresponding to each user respectively according to the user tags;
specifically, label classification can be performed on the transaction flow data and the user data by utilizing natural language processing and clustering to obtain user labels with multiple dimensions, corresponding weights are set for the user labels with the multiple dimensions, and the user portrait is constructed according to the user labels with different dimensions and the weights corresponding to the user labels.
In the present application, each user may be correspondingly constructed with a user representation, which may be associated with at least one business scenario. For example, the user a purchases travel insurance for multiple times, at this time, the user portrait corresponding to the user includes the label of "travel insurance", and the business scene is "travel". In another scenario, user a also purchases car insurance, so the user image corresponding to the user also contains the label "car insurance", and the business scenario is "car". The same user portrait may correspond to one service scenario or to a plurality of service scenarios. The specific service scenario may be pre-selected before prediction by using the bayesian model, which is not limited in the present application.
It should be noted that for a fixed user, a user representation corresponding to the user is formed when the user is marked with a plurality of different labels, and then a user representation model can be constructed from the user representation, so that a user with a certain user representation can be quickly and accurately identified when passing through the user representation model. For users with various types of data being complete, more fine user deep images can be obtained, personalized insurance products are recommended to the users by combining purchase intention values, and response rate and success rate of product recommendation are improved.
Further, the user representation may be updated on-the-fly based on dynamic changes in the user tags. In the present application, the partial user tags may be static and the partial user tags may be dynamic. The dynamic user tags can enable a user to monitor changes of corresponding tag data. Illustratively, feedback of users to bank products and transactions and behavior change characteristics can be identified by setting a plurality of monitoring points, and the like, and the change condition of the dynamic label is detected, so that the user portrait is kept updated immediately, and the effectiveness of user portrait application is enhanced.
It is worth noting that in the application, for the situation that the service scene is simpler, after the user portrait is constructed, the user label in the user portrait can be utilized to carry out comprehensive product promotion, and the user portrait is further perfected by combining the actual situation of the corresponding user, so that more accurate marketing of bank products is carried out.
And step S4: predicting the purchase intention value of each user for the target bank product in a target time range by combining the user portrait based on a Bayesian model;
the purchase intention value can represent quantification of the purchase demand degree of each user for the target bank product. Illustratively, the natural number 0 indicates that the user has no purchase intention, and the natural number 1 indicates that the user has a strong purchase intention, then the purchase intention value may be between 0 and 1. The target time range may include a current or future time instant.
Fig. 3 shows a schematic diagram of a bayesian model of an embodiment of the present application. As shown in fig. 3, the training data of the bayesian model of the present application may be (x), for example 1 ，y 1 )...(x N ，y N ) A set of training parameters is represented. x is the number of 1 -x N Inputs representing N model trainings may include the user's static base data and transaction flow data, y 1 -y N Respectively represent andthe output corresponding to each input can represent the purchase intention value of the user in the historical time range. For example, if a user frequently purchases travel insurance products in 10 months each year from 2018 to 2021, it can be determined that the purchase intention value of the user is 1 in 10 months each year from 2018 to 2021 according to the transaction flow data of the user during 2018 to 2021 and the static basic data. The training data can be input into a Bayesian model for training, the purchase willingness value of the user to the target bank product at the future moment is predicted, and the Bayesian model is continuously corrected according to the dynamic change of the user data.
Further, predicting a purchase intention value of each user for the target bank product in a target time range by combining the user portrait based on a Bayesian model, including:
step S41: based on a Bayesian model, learning and training the associated labels corresponding to the target bank products in the user portrait to obtain probability distribution of a plurality of first model parameters;
for example, in FIG. 3, the output of the Bayesian model may be f θi (. To), wherein f may represent a mapping from training data to a predicted purchase intent value. The first model parameter θ may be derived from training data (x) 1 ，y 1 )...(x N ，y N ) Obtained by middle learning. Notably, the output of the Bayesian model of FIG. 3 can be a set of mappings, e.g., f θi () may represent the ith purchase intention value, i being a natural number. For each of the bayesian model parameters, there is a certain probability of occurrence, which can be represented by P (θ). That is, the ith parameter θ corresponding to the ith output of the Bayesian model of the present application i There may be a mapping with P (θ), and the bayesian model may output a probability distribution of the first model parameter in addition to the purchase intention value.
Step S42: and predicting the purchase intention value of each user for the target bank product in the target time range based on the probability distribution of the plurality of first model parameters.
Compared with the existing deep learning model, the method only uses the slave training data (x) 1 ，y 1 )...(x N ，y N ) The middle learning yields a model with a parameter θ, i.e. f θ To predict the purchase intention of the user, the embodiment of the present application not only outputs the predicted purchase intention value, but also outputs the probability distribution of the bayesian model parameters. Therefore, the Bayesian model of the embodiment of the application is equivalent to a group of models in the prior art to predict the purchase intention value, has the advantage of integrated learning, and can improve the accuracy of the purchase intention value prediction.
Further, predicting a purchase intention value of each user for the target bank product within a target time range based on the probability distribution of the plurality of first model parameters comprises:
step S421: approximating the posterior distribution of the Bayes model by using Gaussian distribution through a variational inference method to obtain estimated distribution;
the variation inference method is also called a variation inference method. Can be used to estimate a function of arbitrary density. In practical applications, a gaussian distribution may be selected to approximate the posterior distribution of the bayesian model, thereby obtaining an estimated distribution corresponding to the objective function.
Step S422: calculating KL divergence between the estimated distribution and the probability distribution of the plurality of first model parameters to obtain the minimized KL divergence;
in particular, in the process of processing the loss function of the bayesian model, since the loss function is dependent on distribution approximation, the embodiment of the present application adopts a variation inference method to minimize the loss function. Specifically, kullback-Liebler (KL) divergence, also known as relative entropy, can be used to measure the distance between different distributions. The loss function of the bayesian model can be expressed by the following formula:
wherein θ ' may represent variation parameters, p (θ ' | D) may represent probability distribution, q (θ ' | μ, σ) may represent estimated distribution, μ and σ may both be parameters in the minimization process, and may be adjusted as required, and D represents the total number of parameters defining the probability distribution.
Step S423: and predicting the purchase intention value of each user for the target bank product in the target time range based on the minimized KL divergence.
By minimizing KL divergence between the estimation distribution and the probability distribution of the plurality of first model parameters, and further minimizing a loss function, the prediction accuracy of the Bayesian model can be further improved.
Further, predicting a purchase intention value of each user for the target bank product in a target time range by combining the user portrait based on a Bayesian model, and further comprising:
step S431: obtaining optimized second model parameters corresponding to the Bayesian model through hyper-parameter search or grid search;
specifically, on the basis of cross validation of the Bayesian model, the hyperparameters in the Bayesian model can be adjusted through a hyperparameter searching method or a grid searching method, so that an optimal hyperparameter set is obtained, and the stability of the Bayesian model is enhanced. The second model parameters may include the hyper-parameters. In practical applications, the initialized trainer clf can be obtained by initializing Bayesian Network (Bayesian Network) parameters, and then learning the training set using clf.fit () function and generating clf instance for training.
Step S432: learning the user data input by the Bayesian model by using the second model parameter to obtain a plurality of historical purchase intention values;
wherein the plurality of historical purchase intention values may be a set of purchase intention values, corresponding to f in FIG. 3 θi (.). Based on the plurality of historical purchase intention values, the purchase intention value of each user for the target bank product in a future target time range can be predicted.
Step S433: and predicting the purchase intention value of each user for the target bank product in the target time range based on the plurality of historical purchase intention values and the probability distribution of the plurality of first model parameters.
Specifically, the test set data can be processed by using a clf.prediction () function and a clf.prediction _ proba () function, then, the probability value and the uncertainty of the user for purchasing the target bank product in the target time range are predicted, and then, the purchase intention value of each user for the target bank product in the target time range is predicted according to the probability value and the uncertainty of the user for purchasing the target bank product in the target time range.
It should be noted that, after prediction is performed by using the bayesian model, further evaluation can be performed on the prediction result. For example, the test set may be predicted by a learned trainer, the purchase probability and the uncertainty of the prediction are obtained, and then the prediction result of at least one bayesian model of accuracy, confusion matrix, receiver Operating Characteristics (ROC) curve, and Stability Index (PSI) evaluation model is evaluated. In the present application, variance may be used to measure uncertainty of the bayesian model for the purchase intention value prediction.
Further, predicting a purchase intention value of each user for the target bank product within a target time range based on the plurality of historical purchase intention values and the probability distribution of the plurality of first model parameters includes:
step S4331: calculating a current variance corresponding to the Bayesian model based on the plurality of historical willingness-to-purchase values;
specifically, the current variance may be calculated according to a set of prediction results of the bayesian model, so as to measure uncertainty of the bayesian model for predicting the purchase intention value, and a formula for calculating the variance is as follows:
wherein f is θ (x) Can represent the prediction result of a Bayesian model, E q (θ) may represent a correlation with a first model parameter θAnd (4) average value. It is noted that there may be more than one prediction, forming a determinant, and before taking the mean value, the matrix and its transpose may be multiplied to obtain the current variance.
Step S4332: judging whether the current variance exceeds a preset variance threshold value;
wherein the variance threshold may be preset. The variance threshold may also be updated and corrected according to a later prediction feedback result, and the size of the variance threshold is not limited in the present application.
Step S4333: if the current variance does not exceed the variance threshold, predicting the purchase intention value of each user for the target bank product within the target time range; and if the current variance exceeds the variance threshold, readjusting the weight parameters of the Bayesian model.
The Bayesian model can be implemented based on a Bayesian neural network. The bayesian neural network is different from a general neural network in that a weight parameter of the bayesian neural network is a random variable rather than a definite value.
Fig. 4 shows a schematic diagram of a bayesian neural network of an embodiment of the present application. As shown in FIG. 4, X denotes an input layer, Y denotes an output layer, Z 1 And Z 2 Are all intermediate layers of Bayesian neural networks. θ 1, θ 2, θ 3 may be weight parameters assigned to user tags in corresponding user representations during neural network training. These weight parameters may each vary dynamically based on probability. The weight parameters are combined with the neural network through probability modeling, the confidence degree of a prediction result can be given, and the prior can also be used for describing key parameters and used as the input of the neural network.
The uncertainty of prediction is analyzed by calculating the current variance according to a group of prediction results of the Bayesian model, and the user images are combined, so that more targeted and accurate marketing can be provided.
Step S5: and recommending the target bank product to a target user according to the purchase intention value.
Specifically, recommending the target bank product to the target user according to the purchase intention value may include:
step S51: comparing the purchase intention value with a preset purchase intention threshold value, and generating a target user white list according to a comparison result;
in one example, the preset purchase intention threshold may be 0.8, and the probability that the user a purchases the travel insurance product in the future 2023 years is 1, which is obtained through the bayesian model, and is greater than the purchase intention threshold, so that the user may be determined as a user with high purchase intention corresponding to the travel insurance product, and the user may be added to the white list of the corresponding target users. The target user white list can be further subdivided according to different application scenarios, and the specific content of the white list is not limited in the application.
Step S52: and recommending the target bank product to the target user according to the white list of the target user.
For example, after a user manager of a bank obtains a highly willing user in a white list of a target user, a marketing object and a marketing opportunity can be selected according to the white list to develop insurance-added marketing or change a potential insurance user into a business of a real insurance user. In practical application, the white list can be issued to offline personnel and further analyzed. In addition, the response and feedback of the user can be recorded in a data system for user tag maintenance and update.
In summary, with the help of a data platform, according to user data, the embodiment of the application can predict the purchase intention value of each user for the target bank product within the target time range by combining a Bayesian model, and then recommend the target bank product to the target user more accurately. On one hand, the accurate marketing of the bank products can predict the behavior mode of the bank products according to various habits of users, and the selling success rate of the target bank products is improved. On the other hand, compare with modes such as traditional marketing, telemarketing even sweep street marketing, the accurate marketing of bank product can practice thrift a large amount of manpower and materials, improves the accurate degree of marketing, has practiced thrift a large amount of marketing costs. Moreover, various targeted insurance information is actively pushed to potential users or target users by accurately analyzing user requirements, and ways such as telemarketing, visiting and configuring rights and interests activities are pushed to wake up sleeping users, so that the user stickiness can be increased, and the marketing efficiency is higher. In addition, compared with a general deep learning method, the deep learning method is designed under a Bayesian inference framework, inherits the interpretability of a Bayesian statistical theory, and is safer and more reliable.
In order to better implement the method, correspondingly, the embodiment of the application further provides an intelligent bank product recommending device, and the intelligent bank product recommending device is specifically integrated in a terminal or a server.
Fig. 5 shows a schematic diagram of an intelligent bank product recommendation device according to an embodiment of the present application. Referring to fig. 5, the apparatus includes:
the data acquisition module 51 is configured to acquire user data of a target bank product within a preset historical time range, where the user data includes static basic data related to each user and transaction flow data formed by each user in a transaction process;
a tag generation module 52, configured to perform classification processing on the user data, and generate a multi-dimensional user tag corresponding to each user;
the portrait construction module 53 is configured to construct, according to the user tag, a user portrait of at least one service scene corresponding to each user;
the prediction module 54 is used for predicting the purchase intention value of each user for the target bank product by combining the user portrait based on a Bayesian model;
and the recommending module 55 is configured to recommend the target bank product to the target user according to the purchase intention value.
In addition, an embodiment of the present application further provides a computer device, where the computer device may be a terminal or a server, as shown in fig. 6, which shows a schematic structural diagram of the computer device according to the embodiment of the present application, and specifically:
the computer device may include components such as a processor 401 of one or more processing cores, memory 402 of one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will appreciate that the computer device configuration illustrated in FIG. 6 does not constitute a limitation of computer devices, and may include more or fewer components than those illustrated, or some components may be combined, or a different arrangement of components. Wherein:
the processor 401 is a control center of the computer device, connects various parts of the entire computer device using various interfaces and lines, and performs various functions of the computer device and processes data by running or executing software programs and/or modules stored in the memory 402 and calling data stored in the memory 402, thereby monitoring the computer device as a whole. Optionally, processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor, which mainly handles operating systems, user interfaces, application programs, etc., and a modem processor, which mainly handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 401.
The memory 402 may be used to store software programs and modules, and the processor 401 executes various functional applications and data processing by operating the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created according to use of the computer device, and the like. Further, the memory 402 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 access to the memory 402.
The computer device further comprises a power supply 403 for supplying power to the various components, and preferably, the power supply 403 is logically connected to the processor 401 via a power management system, so that functions of managing charging, discharging, and power consumption are implemented via the power management system. The power supply 403 may also include any component of one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.
The computer device may also include an input unit 404, the input unit 404 being operable to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.
Although not shown, the computer device may further include a display unit and the like, which are not described in detail herein. Specifically, in this embodiment, the processor 401 in the computer device loads the executable file corresponding to the process of one or more application programs into the memory 402 according to the following instructions, and the processor 401 runs the application programs stored in the memory 402, thereby implementing various functions as follows:
step S1: acquiring user data of a target bank product in a preset historical time range, wherein the user data comprises static basic data related to each user and transaction flow data formed by each user in a transaction process;
step S2: classifying the user data to generate multi-dimensional user tags respectively corresponding to the users;
and step S3: constructing user images of at least one service scene corresponding to each user respectively according to the user tags;
and step S4: predicting the purchase intention value of each user for the target bank product in a target time range by combining the user portrait based on a Bayesian model;
step S5: and recommending the target bank product to a target user according to the purchase intention value.
The above operations can be implemented in the foregoing embodiments, and are not described in detail herein.
It will be understood by those skilled in the art that all or part of the steps of the methods of the above embodiments may be performed by instructions or by associated hardware controlled by the instructions, which may be stored in a computer readable storage medium and loaded and executed by a processor.
To this end, the embodiments of the present application also provide a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the method provided in various optional implementation manners in the embodiments of the present application.
According to an aspect of the application, there is also provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions to cause the computer device to perform the method provided in the various alternative implementations in the embodiments described above.
The above operations can be implemented in the foregoing embodiments, and are not described in detail herein.
Wherein the storage medium may include: read Only Memory (ROM), random Access Memory (RAM), magnetic or optical disks, and the like.
Since the instructions stored in the storage medium can execute the steps in the method provided by the embodiment of the present application, the beneficial effects that can be achieved by the method provided by the embodiment of the present application can be achieved, for details, see the foregoing embodiments, and are not described herein again.
The method, the apparatus, the computer device and the storage medium for intelligently recommending bank products provided by the embodiment of the present application are introduced in detail, a specific example is applied in the present application to explain the principle and the implementation of the present application, and the description of the above embodiment is only used to help understand the method and the core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Claims (10)
1. A method for intelligently recommending bank products, the method comprising:
acquiring user data of a target bank product in a preset historical time range, wherein the user data comprises static basic data related to each user and transaction flow data formed by each user in a transaction process;
classifying the user data to generate multi-dimensional user tags respectively corresponding to the users;
constructing user images of at least one service scene corresponding to each user respectively according to the user tags;
predicting the purchase intention value of each user for the target bank product in a target time range by combining the user portrait based on a Bayesian model;
and recommending the target bank product to a target user according to the purchase intention value.
2. The method for intelligently recommending bank products according to claim 1, wherein predicting the value of the purchase intention of each user for said target bank product within the target time range based on the bayesian model in combination with said user representation comprises:
based on a Bayesian model, learning and training the associated labels corresponding to the target bank products in the user portrait to obtain probability distribution of a plurality of first model parameters;
and predicting the purchase intention value of each user for the target bank product in the target time range based on the probability distribution of the plurality of first model parameters.
3. The method for intelligently recommending bank products according to claim 2, wherein predicting the value of the purchase intention of each user for the target bank product within the target time range based on the probability distribution of the plurality of first model parameters comprises:
approximating the posterior distribution of the Bayes model by using Gaussian distribution through a variational inference method to obtain estimated distribution;
calculating KL divergence between the estimated distribution and the probability distribution of the plurality of first model parameters to obtain minimized KL divergence;
and predicting the purchase intention value of each user for the target bank product in the target time range based on the minimized KL divergence.
4. The method for intelligently recommending bank products according to claim 2, wherein predicting the value of the purchase intention of each user for said target bank product within the target time range based on the bayesian model in combination with said user representation further comprises:
obtaining optimized second model parameters corresponding to the Bayesian model through hyper-parameter search or grid search;
learning the user data input by the Bayesian model by using the second model parameter to obtain a plurality of historical purchase intention values;
and predicting the purchase intention value of each user for the target bank product in the target time range based on the plurality of historical purchase intention values and the probability distribution of the plurality of first model parameters.
5. The method for intelligently recommending bank products according to claim 4, wherein predicting the purchase intention value of each user for said target bank product within the target time range based on said plurality of historical purchase intention values and the probability distribution of said plurality of first model parameters comprises:
calculating a current variance corresponding to the Bayesian model based on the plurality of historical willingness-to-purchase values;
judging whether the current variance exceeds a preset variance threshold value;
if the current variance does not exceed the variance threshold, predicting the purchase intention value of each user for the target bank product within the target time range; and if the current variance exceeds the variance threshold, readjusting the weight parameters of the Bayesian model.
6. The method according to claim 1, wherein the classifying the user data to generate multi-dimensional user tags respectively corresponding to the users comprises:
preprocessing the user data to obtain preprocessed user data;
and performing classification processing according to the preprocessed user data to generate multi-dimensional user tags respectively corresponding to the users, wherein the dimensions of the user tags at least comprise tag themes, tag sources and tag attributes.
7. The method for intelligently recommending bank products according to claim 1, wherein recommending the target bank product to the target user according to the purchase intention value comprises:
comparing the purchase intention value with a preset purchase intention threshold value, and generating a target user white list according to a comparison result;
and recommending the target bank product to the target user according to the target user white list.
8. An intelligent bank product recommendation device, comprising:
the data acquisition module is used for acquiring user data of a target bank product within a preset historical time range, wherein the user data comprises static basic data related to each user and transaction flow data formed by each user in a transaction process;
the label generation module is used for classifying the user data and generating multi-dimensional user labels respectively corresponding to the users;
the image construction module is used for constructing user images of at least one service scene respectively corresponding to the users according to the user tags;
the prediction module is used for predicting the purchase intention value of each user for the target bank product by combining the user portrait based on a Bayesian model;
and the recommending module is used for recommending the target bank product to the target user according to the purchase intention value.
9. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when the computer program is run on a computer, causes the computer to carry out the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211371022.XA CN115631006A (en) | 2022-11-03 | 2022-11-03 | Method and device for intelligently recommending bank products, storage medium and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211371022.XA CN115631006A (en) | 2022-11-03 | 2022-11-03 | Method and device for intelligently recommending bank products, storage medium and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115631006A true CN115631006A (en) | 2023-01-20 |
Family
ID=84908946
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211371022.XA Pending CN115631006A (en) | 2022-11-03 | 2022-11-03 | Method and device for intelligently recommending bank products, storage medium and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115631006A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116228278A (en) * | 2023-03-10 | 2023-06-06 | 读书郎教育科技有限公司 | User portrait establishing method and user portrait management system based on big data |
-
2022
- 2022-11-03 CN CN202211371022.XA patent/CN115631006A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116228278A (en) * | 2023-03-10 | 2023-06-06 | 读书郎教育科技有限公司 | User portrait establishing method and user portrait management system based on big data |
| CN116228278B (en) * | 2023-03-10 | 2023-11-14 | 读书郎教育科技有限公司 | User portrait establishing method and user portrait management system based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8719192B2 (en) | Transfer of learning for query classification | |
| US9355067B1 (en) | Distribution of parameter calculation for iterative optimization methods | |
| KR102404511B1 (en) | Method, device and system for recommending customized product information based on artificial intelligence | |
| CN111274330B (en) | Target object determination method and device, computer equipment and storage medium | |
| CN115002200B (en) | Message pushing method, device, equipment and storage medium based on user portrait | |
| WO2021155691A1 (en) | User portrait generating method and apparatus, storage medium, and device | |
| CN110598070A (en) | Application type identification method and device, server and storage medium | |
| US20230315618A1 (en) | Data processing method and device, computing device, and test reduction device | |
| CN111429161B (en) | Feature extraction method, feature extraction device, storage medium and electronic equipment | |
| CN112633690A (en) | Service personnel information distribution method, service personnel information distribution device, computer equipment and storage medium | |
| Yakhchi et al. | CNR: cross-network recommendation embedding user’s personality | |
| CN113672797A (en) | Content recommendation method and device | |
| Khan et al. | Comparative analysis on Facebook post interaction using DNN, ELM and LSTM | |
| CN117271767A (en) | Operation and maintenance knowledge base establishing method based on multiple intelligent agents | |
| CN115631006A (en) | Method and device for intelligently recommending bank products, storage medium and computer equipment | |
| CN115730125A (en) | Object identification method and device, computer equipment and storage medium | |
| US20230308360A1 (en) | Methods and systems for dynamic re-clustering of nodes in computer networks using machine learning models | |
| CN117235264A (en) | Text processing method, apparatus, device and computer readable storage medium | |
| Yin et al. | Rsygan: Generative adversarial network for recommender systems | |
| US20160042277A1 (en) | Social action and social tie prediction | |
| CN112200602A (en) | Neural network model training method and device for advertisement recommendation | |
| CN117009883B (en) | Object classification model construction method, object classification method, device and equipment | |
| Zhang et al. | Temporal dynamic recommendation based on data imputation through association analysis | |
| EP4354340A1 (en) | Translation decision assistant | |
| Cao et al. | Parameter optimization of simulation models for material supply in an emergent disaster based on support vector machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |