CN115578165A - Product recommendation method and device for financial institution, electronic device and storage medium - Google Patents
Product recommendation method and device for financial institution, electronic device and storage medium Download PDFInfo
- Publication number
- CN115578165A CN115578165A CN202211303505.6A CN202211303505A CN115578165A CN 115578165 A CN115578165 A CN 115578165A CN 202211303505 A CN202211303505 A CN 202211303505A CN 115578165 A CN115578165 A CN 115578165A
- Authority
- CN
- China
- Prior art keywords
- product
- sample
- label
- training
- user
- 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
- 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
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
-
- 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
- G06Q40/06—Asset management; Financial planning or analysis
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Technology Law (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a product recommendation method and device for a financial institution, electronic equipment and a storage medium, and relates to the field of artificial intelligence, wherein the method comprises the following steps: receiving user data, wherein the user data comprises: user characteristics, transaction detail data and transaction behaviors; inputting user data into a product recommendation model, and receiving a recommendation result output by the product recommendation model, wherein training data of the product recommendation model is integrated with label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and a product level to which each quasi-recommended product belongs; and displaying the recommendation result on a system page. The invention solves the technical problem that the recommendation effect is poor because the hierarchical relationship among products is ignored when the financial products are recommended by the recommendation model in the prior art.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to a product recommendation method and device for a financial institution, electronic equipment and a storage medium.
Background
With the development of science and technology, financial institutions have created online and offline rich and diverse user contacts in order to meet the requirements of users for daily business handling, channel transaction and the like. Facing a large number of users, financial institutions need to more comprehensively and accurately address financial management needs of users. In the actual process of carrying out the business of the financial products, the attractiveness of different financial products to the customer base needs to be mined, so that a target customer base is found and targeted marketing is carried out.
In the related art, the mainstream machine learning technology has obvious defects/shortcomings in the precise marketing scene of a customer base, which are mainly embodied in two aspects, namely, firstly, only one product can be modeled in the related art, and if a user needs to predict whether to purchase a plurality of products, each product needs to be modeled and learned, so that the model construction cost is too high, and each model needs to be constructed with a training sample independently. Secondly, in the related art, even if some modeling methods support modeling of all products, attributes of the products are generally used as product features to be added into model learning so as to distinguish different products, and hierarchical relationships among the products are ignored, so that the learning effect is poor.
In view of the above problems, no effective solution has been proposed.
Disclosure of Invention
The embodiment of the invention provides a product recommendation method and device for a financial institution, electronic equipment and a storage medium, which are used for at least solving the technical problem that the recommendation effect is poor due to the fact that the hierarchical relationship among products is ignored when a recommendation model in the prior art recommends financial products.
According to an aspect of an embodiment of the present invention, there is provided a product recommendation method of a financial institution, including: receiving user data, wherein the user data comprises: user characteristics, transaction detail data and transaction behaviors; inputting the user data into a product recommendation model, and receiving a recommendation result output by the product recommendation model, wherein training data of the product recommendation model is integrated with label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and the product level to which each quasi-recommended product belongs; and displaying the recommendation result on a system page.
Optionally, when building the product recommendation model, the method includes: obtaining a plurality of training samples, wherein each training sample at least comprises: user characteristics, historical transaction behaviors, and historical transaction detail data; constructing sample characteristics of each training sample; determining a hierarchical multi-label experience loss term and a hierarchical multi-label comparison loss term corresponding to each training sample based on the sample characteristics; determining a model objective function based on the hierarchical multi-label empirical loss term and the hierarchical multi-label comparative loss term; and adjusting the model objective function to obtain a classifier of each product level, and determining the product recommendation model.
Optionally, after obtaining the plurality of training samples, further comprising: analyzing data columns of a data table in each training sample; completing missing values in the data columns; or deleting the data columns with the total number of the missing items to which the missing values belong being larger than a preset number threshold.
Optionally, the step of constructing a sample feature of each training sample includes: extracting basic information related to each user in each training sample to obtain user characteristics of the user; extracting behavior preference information associated with each user in each training sample to obtain behavior characteristics; extracting basic information of related transaction products in each training sample to obtain product characteristics; extracting transaction data related to the transaction product in each training sample to obtain product transaction characteristics; and constructing sample characteristics of each training sample based on the user characteristics, the behavior characteristics, the product characteristics and the product transaction characteristics.
Optionally, the step of determining a hierarchical multi-label experience loss term and a hierarchical multi-label contrast loss term corresponding to each training sample based on the sample features includes: determining a plurality of sample labels for the training sample based on the sample features; configuring a plurality of the product levels corresponding to trading products of the training sample, and configuring the corresponding classifier for each product level; determining, by the classifier, a product hierarchy to which the training sample belongs based on the plurality of sample labels; establishing a sample label hierarchical model based on the sample set to which the training sample belongs, the plurality of sample labels, the product level to which the training sample belongs and the classifier corresponding to the product level, and determining a hierarchical multi-label experience loss term of the training sample by the sample label hierarchical model.
Optionally, the hierarchical multi-label experience loss item is used for characterizing hierarchical information of a product hierarchy to which the transaction product belongs and hierarchical association relations.
Optionally, the step of determining a hierarchical multi-label experience loss term and a hierarchical multi-label contrast loss term corresponding to each training sample based on the sample features further includes: pairing all the training samples to construct a plurality of training sample pairs; inquiring a sample label corresponding to each training sample in the training sample pair and a product level to which the sample label belongs; characterizing the training sample pairs belonging to the same product level and having the same sample label as a positive sample pair; characterizing the training sample pairs which do not belong to the same product level and/or the sample labels are not the same as a negative sample pair; adopting a preset comparison learning strategy to output the generalization distance between the two training samples related to the positive sample pair in the sample set to be close and output the generalization distance between the two training samples related to the negative sample pair to be far; and controlling the product recommendation model to learn similar information between the training samples in the positive sample pair, learning distinguishing information between the training samples in the negative sample pair, and determining a hierarchical multi-label contrast loss item of the training samples by the product recommendation model.
According to another aspect of the embodiments of the present invention, there is also provided a product recommendation apparatus for a financial institution, including: a receiving unit, configured to receive user data, where the user data includes: user characteristics, transaction detail data and transaction behaviors; the input unit is used for inputting the user data into a product recommendation model and receiving a recommendation result output by the product recommendation model, wherein the training data of the product recommendation model is integrated with label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and the product level to which each quasi-recommended product belongs; and the display unit is used for displaying the recommendation result on a system page.
Optionally, the product recommendation device of the financial institution further comprises: a first obtaining subunit, configured to obtain a plurality of training samples, where each training sample at least includes: user characteristics, historical transaction behaviors, and historical transaction detail data; the first construction subunit is used for constructing the sample characteristics of each training sample; the first determining subunit is used for determining a hierarchical multi-label experience loss item and a hierarchical multi-label comparison loss item corresponding to each training sample based on the sample characteristics; a second determining subunit, configured to determine a model objective function based on the hierarchical multi-label empirical loss term and the hierarchical multi-label comparison loss term; and the first adjusting subunit is used for adjusting the model objective function to obtain the classifier of each product level and determine the product recommendation model.
Optionally, the first obtaining subunit includes: the first analysis module is used for analyzing data columns of the data table in each training sample; the first completion module is used for performing completion processing on missing values in the data columns; or the first deleting module is used for deleting the data columns with the total number of the missing items of which the missing values belong to being larger than a preset number threshold.
Optionally, the first building subunit comprises: the first extraction module is used for extracting basic information related to each user in each training sample to obtain user characteristics of the user; the second extraction module is used for extracting behavior preference information associated with each user in each training sample to obtain behavior characteristics; the third extraction module is used for extracting basic information of related transaction products in each training sample to obtain product characteristics; the fourth extraction module is used for extracting the transaction data related to the transaction product in each training sample to obtain product transaction characteristics; a first constructing module, configured to construct a sample feature of each training sample based on the user feature, the behavior feature, the product feature, and the product transaction feature.
Optionally, the first determining subunit includes: a first determination module to determine a plurality of sample labels of the training sample based on the sample features; a first configuration module, configured to configure a plurality of product levels corresponding to the trading products of the training sample, and configure a corresponding classifier for each of the product levels; a second determination module, configured to determine, by the classifier, a product hierarchy to which the training sample belongs based on the plurality of sample labels; the first establishing module is used for establishing a sample label hierarchical model based on the sample set to which the training sample belongs, the plurality of sample labels, the product level to which the training sample belongs and the classifier corresponding to the product level, and determining a hierarchical multi-label experience loss term of the training sample by the sample label hierarchical model.
Optionally, the hierarchical multi-label experience loss item is used for characterizing hierarchical information of a product hierarchy to which the transaction product belongs and hierarchical association relations.
Optionally, the first determining subunit further includes: the second construction module is used for pairing all the training samples to construct a plurality of training sample pairs; the first query module is used for querying the sample label and the product level corresponding to each training sample in the training sample pair; the first characterization module is used for characterizing the training sample pairs which belong to the same product level and have the same sample labels as positive sample pairs; the second characterization module is used for characterizing the training sample pairs which do not belong to the same product level and/or have different sample labels as negative sample pairs; the first output module is used for outputting the generalization distance between the two training samples related to the positive sample pair in the sample set to be close and outputting the generalization distance between the two training samples related to the negative sample pair to be far by adopting a preset comparison learning strategy; and the first learning module is used for controlling a product recommendation model to learn similar information between the training samples in the positive sample pair, learning different information between the training samples in the negative sample pair, and determining a hierarchical multi-label contrast loss item of the training samples by the product recommendation model.
According to another aspect of the embodiments of the present invention, there is further provided a computer-readable storage medium including a stored computer program, wherein when the computer program runs, the apparatus on which the computer-readable storage medium is located is controlled to execute the method for recommending a product by a financial institution.
According to another aspect of embodiments of the present invention, there is also provided an electronic device, including one or more processors and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors are caused to implement the product recommendation method for a financial institution of any one of the above.
In the present disclosure, the following steps are employed: receiving user data, inputting the user data into a product recommendation model, receiving a recommendation result output by the product recommendation model, and displaying the recommendation result on a system page, wherein training data of the product recommendation model is fused into label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and a product level to which each quasi-recommended product belongs.
According to the method and the device, label information of multiple product levels is integrated into a product recommendation model, attributes of different levels of products are used as product recommendation bases, diversity and accuracy of product recommendation are improved, user requirements are fully met, user experience is improved, and therefore the technical problem that when financial products are recommended by the recommendation model in the prior art, hierarchical relations among the products are ignored, and the recommendation effect is poor is solved.
In the method, the layered multi-label comparison learning items are adopted, so that similar product sample pairs are closer to each other in the label space of each level, the sample pairs with large differences are farther away from each other in the label space of each level, and the generalization performance of the model is improved.
In the method, a big data technology is utilized, user information (such as user behaviors, asset information and product transaction information) is used for machine learning modeling, the probability of purchasing various financial products by a user is predicted, the prediction result is applied to financial marketing, efficient recommendation of the financial products is achieved, and the method has important significance for meeting financial requirements of the user.
According to the method and the system, key marketing groups can be selected according to the prediction result of the product recommendation model, marketing strength is enhanced, business trends of different users are excavated, a product marketing scheme is provided in a targeted mode, marketing success probability is increased, and a more targeted solution is provided for service upgrading of financial institutions.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention and do not constitute a limitation of the invention. In the drawings:
FIG. 1 is a flow chart of an alternative financial institution product recommendation method according to embodiments of the invention;
FIG. 2 is a flow diagram of an alternative method of building a product recommendation model according to embodiments of the invention;
FIG. 3 is a diagram illustrating an alternative product recommendation architecture for a financial institution in accordance with an embodiment of the present invention;
FIG. 4 is a schematic diagram of an alternative financial institution product recommendation device in accordance with embodiments of the invention;
fig. 5 is a block diagram of a hardware configuration of an electronic device (or a mobile device) of a product recommendation method of a financial institution according to an embodiment of the present invention.
Detailed Description
In order to make those skilled in the art better understand the technical solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
It should be noted that the product recommendation method and apparatus of the financial institution in the present disclosure may be used in the field of artificial intelligence, and may also be used in any field other than the field of artificial intelligence under the condition of recommending financial products to users/generating financial product recommendation schemes according to multi-product hierarchical labels, and the application fields of the product recommendation method and apparatus of the financial institution in the present disclosure are not limited.
It should be noted that the relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for presentation, analyzed data, etc.) referred to in the present disclosure are information and data authorized by the user or sufficiently authorized by each party. For example, an interface is provided between the system and the relevant user or organization, before obtaining the relevant information, an obtaining request needs to be sent to the user or organization through the interface, and after receiving the consent information fed back by the user or organization, the relevant information is obtained.
The method can be applied to various product recommendation software/devices/products (such as various mobile terminals, PC terminals and financial institutions APP), label information of multiple product levels is merged into a product recommendation model, attributes of different levels of products are used as product recommendation bases, diversity and accuracy of recommended products are improved, user requirements are fully met, and user experience is improved.
The financial products related to the invention have a hierarchical relationship, each financial product (or financial product) has different levels of category attributes, for example, deposit products comprise fixed periods and current periods, wherein the fixed period deposit products comprise large fixed periods, 5-year fixed periods, 3-year fixed periods and the like, and the expressed product hierarchy is downward; or 3-year regular products of the products belong to regular products on the upper layer and the RMB deposit products on the upper layer, and the product is represented in an upward layer; relationships between different classes are effectively represented through product hierarchies (or expressed as product hierarchies), and each product can be described from different hierarchies respectively, so that label information of multiple hierarchies can be fully utilized to optimize a machine learning model. In the product recommendation model in the related art, only the product attribute is used as the product characteristic, the hierarchical relation is ignored, and the product hierarchy is applied to model learning in the invention, so that the learning effect is improved.
The present invention will be described in detail with reference to examples.
Example one
In accordance with an embodiment of the present invention, there is provided a method embodiment for product recommendation by a financial institution, it is noted that the steps illustrated in the flowchart of the drawings may be performed in a computer system such as a set of computer executable instructions, and that while a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than here.
Fig. 1 is a flowchart of a product recommendation method of an alternative financial institution according to an embodiment of the present invention, as shown in fig. 1, the method includes the following steps:
step S101, receiving user data, wherein the user data comprises: user characteristics, transaction detail data and transaction behaviors;
step S102, inputting user data into a product recommendation model, and receiving a recommendation result output by the product recommendation model, wherein training data of the product recommendation model is integrated into label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and a product level to which each quasi-recommended product belongs;
and step S103, displaying the recommendation result on a system page.
Through the steps, the user data are received firstly, then the user data are input into the product recommendation model, the recommendation result output by the product recommendation model is received, and finally the recommendation result is displayed on a system page, wherein the training data of the product recommendation model is fused into label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and a product level to which each quasi-recommended product belongs. In this embodiment, the label information of a plurality of product levels is merged into the product recommendation model, the attributes of different levels of the product are used as the basis for product recommendation, the diversity and accuracy of product recommendation are improved, the user requirements are fully met, the user experience is improved, and the technical problem that the recommendation effect is poor due to the fact that the hierarchical relationship between the products is ignored when the financial products are recommended by the recommendation model in the prior art is solved.
The following will explain each implementation procedure in detail.
Step S101, receiving user data, wherein the user data comprises: user characteristics, transaction detail data, and transaction behavior.
User characteristics include, but are not limited to: user gender, user age (which may be expressed by age group or age-specific years), user occupation (which may be expressed by position, company operating product field, etc., or by specific occupation name), transaction detail data for characterizing the user's bill details for purchasing financial products over a historical period of time, transaction activities including, but not limited to: the transaction type of the user and the configuration situation of the financial product.
Step S102, inputting user data into a product recommendation model, and receiving a recommendation result output by the product recommendation model, wherein training data of the product recommendation model is integrated into label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and a product level to which each quasi-recommended product belongs.
It should be noted that before inputting the user data into the product recommendation model, the product recommendation model needs to be trained first. FIG. 2 is a flow chart of an alternative method for building a product recommendation model according to an embodiment of the present invention, and as shown in FIG. 2, the step of building the product recommendation model includes:
step S201, obtaining a plurality of training samples, wherein each training sample at least includes: user characteristics, historical transaction behaviors and historical transaction detail data;
it should be noted that, before training the product recommendation model, various information data of the user needs to be acquired, the information data of the user is fully utilized, a plurality of training samples and test samples are generated, the training speed of the model and the model prediction accuracy, such as user characteristics, user transaction detail data, user transaction behaviors and the like, can be improved through the effective training samples and test samples, so that the probability of purchasing various financial products by the user is predicted, the prediction result is applied to a marketing scheme, the user demand can be better known, and the financial products are accurately provided for the user.
In the embodiment of the present invention, after obtaining a plurality of training samples, data in the training samples may be preprocessed to obtain more complete sample data, and during the preprocessing, the preprocessing may include: analyzing the data columns of the data table in each training sample; completing missing values in the data columns; or deleting the data columns with the total number of the missing items of which the missing values belong to being larger than a preset number threshold.
Step S202, sample characteristics of each training sample are constructed.
In this embodiment of the present invention, step S202 may include: extracting basic information related to each user in each training sample to obtain user characteristics of the user; extracting behavior preference information associated with each user in each training sample to obtain behavior characteristics; extracting basic information of related transaction products in each training sample to obtain product characteristics; extracting transaction data of related transaction products in each training sample to obtain product transaction characteristics; and constructing sample characteristics of each training sample based on the user characteristics, the behavior characteristics, the product characteristics and the product transaction characteristics.
It should be noted that the product recommendation model may be learned based on user information for different users, so as to recommend a financial product meeting a user expectation value for a specific user, and therefore, when performing model training, corresponding sample features need to be constructed based on basic information, user behavior information, product information, and product transaction information of the specific user. Extracting basic information of the user as user characteristics, such as the age, sex, occupation and other basic characteristics of the user (corresponding to the user characteristics); extracting behavior preferences of the user, such as transaction category distribution, financial product configuration condition and the like, as behavior characteristics of the user; extracting basic information of the product as product characteristics; and extracting transaction data of the product, such as transaction times, transaction modes and the like, as product transaction characteristics.
Step S203, determining a layered multi-label experience loss item and a layered multi-label comparison loss item corresponding to each training sample based on the sample characteristics.
In the embodiment of the invention, the step of determining the hierarchical multi-label experience loss item and the hierarchical multi-label comparison loss item corresponding to each training sample comprises the following steps: determining a plurality of sample labels of the training sample based on the sample features; configuring a plurality of product levels corresponding to the transaction products of the training samples, and configuring a corresponding classifier for each product level; determining, by a classifier, a product hierarchy to which a training sample belongs based on a plurality of sample labels; establishing a sample label hierarchical model based on a sample set to which a training sample belongs, a plurality of sample labels, a product level to which the training sample belongs and a classifier corresponding to the product level, and determining a hierarchical multi-label experience loss term of the training sample by the sample label hierarchical model.
In the embodiment of the invention, the hierarchical multi-label experience loss item is used for representing the hierarchical information of the product hierarchy to which the transaction product belongs and the hierarchical incidence relation.
It should be noted that, a hierarchical multi-label experience loss item is obtained, so that the label information of multiple levels can be fused and learned in the optimization process of the machine learning model, and the generalization effect of the model is improved. The calculation formula of the hierarchical multi-label empirical loss term is as follows:
wherein, X is a training sample set, X is a certain training sample therein, L is a multi-layer label of the training sample (namely determining the product level to which the training sample belongs), L is a certain layer sample label therein, f l () Classifier corresponding to the L-th label, L emp Is a tag experience loss term.
The embodiment improves the past experience loss items, which are called hierarchical multi-label experience loss items, and the model can fully learn the hierarchical information of the product.
In this embodiment of the present invention, when acquiring the hierarchical multi-label comparison loss item, step S203 further includes: pairing all training samples to construct a plurality of training sample pairs; inquiring a sample label corresponding to each training sample in the training sample pair and the product level to which the sample label belongs; characterizing training sample pairs which belong to the same product level and have the same sample label as a positive sample pair; representing training sample pairs which do not belong to the same product level and/or have different sample labels as negative sample pairs; adopting a preset comparison learning strategy to output the generalization distance between two training samples related to the positive sample pair in the sample set to be close and output the generalization distance between two training samples related to the negative sample pair to be far; and controlling the product recommendation model to learn similar information between the training samples in the positive sample pair, learning distinguishing information between the training samples in the negative sample pair, and determining the hierarchical multi-label contrast loss item of the training samples by the product recommendation model.
It should be noted that, when obtaining the hierarchical multi-label comparison loss item, in this embodiment, a training sample pair needs to be constructed first, if two training samples are on the l-th label and have the same label information, that is, two users both hold a certain product, the two training samples are positive sample pairs on the l-th label (corresponding to the above-mentioned training sample pair that belongs to the same product level and has the same sample label being characterized as a positive sample pair), otherwise, the two training samples are negative sample pairs (corresponding to the above-mentioned training sample pair that does not belong to the same product level and/or has different sample labels being characterized as a negative sample pair). In this embodiment, when the contrast loss value of the training sample pair is calculated, the calculation formula adopted is as follows:
wherein X and X' represent two training samples (corresponding to the above sample pair), X is the training sample set including all the training samples, f l () Is a classifier corresponding to the label of the L-th layer, a is other training samples different from x, tau is a hyperparameter, L pair (x, x') is the contrast loss value of the two training samples.
It should be noted that, in the process of optimizing the model, training sample pairs are selected for each layer of sample labels and learning training is performed to obtain a hierarchical multi-label comparison loss item, in the process, the model learns similar information between the training samples in the positive sample pairs and learns distinguishing information between the training samples in the negative sample pairs, so that the positive sample pairs can be close to each other, the negative sample pairs can be far away from each other, and through comparison learning, the model fully learns the distinguishing information between the similar information between the positive samples and the distinguishing information between the negative samples, and the generalization effect of the model is improved.
Wherein, the following formula is adopted to calculate the layering multi-label contrast loss term:
wherein L is a product level, L is an L-th level sample label, X is a training sample set, X is a training sample in an X indication sample set, and P is l x To form a sample set of positive sample pairs with the training sample x,
is P l x One sample of the indicated set of samples;
for a set of samples that form negative sample pairs with sample x,
is composed of
One sample of the indicated set of samples;
represents sample x and sample
The value of the loss of contrast between,
represents sample x and sample
Loss of contrast between, L hmc The lossy terms are compared for hierarchical multi-labels.
In the embodiment, in order to enable the model to fully learn the purchase information of the previous user/customer on financial products (or financial products) and make preferred product recommendation to the customer, a hierarchical multi-label comparison learning item is designed, so that similar sample pairs are closer to each other in the label space of each product level, and sample pairs with large differences are farther from each other in the label space of each level, thereby improving the generalization performance of the model.
And S204, determining a model objective function based on the hierarchical multi-label empirical loss term and the hierarchical multi-label comparison loss term.
It should be noted that, after obtaining the hierarchical multi-label experience loss term and the hierarchical multi-label comparison loss term, the final objective function (corresponding to the model objective function described above) of the hierarchical multi-label product recommendation model based on the comparison learning is as follows:
Loss＝L emp +L hmc ；
wherein Loss is an objective function, L emp For hierarchical multi-label empirical loss terms, L hmc The lossy terms are compared for hierarchical multi-labels.
And S205, adjusting a model objective function to obtain a classifier of each product level, and determining a product recommendation model.
Through the optimization of the objective function, the final classifier f of each product level is obtained l () And inputting the x to the trained classifier for the sample x to be tested, so as to obtain the prediction result of each product level.
And step S103, displaying the recommendation result on a system page.
Through the embodiment, the label information of multiple product levels is merged into the product recommendation model, the attributes of different levels of the product serve as the product recommendation basis, the diversity and accuracy of product recommendation are improved, the user requirements are fully met, the user experience is improved, and the technical problem that the recommendation effect is poor due to the fact that the hierarchical relation between the products is ignored when financial products are recommended by the recommendation model in the prior art is solved.
The invention is illustrated below with reference to another detailed embodiment.
FIG. 3 is an architecture diagram of an alternative product recommendation model for a financial institution according to an embodiment of the invention, and as shown in FIG. 3, the training of the product recommendation model for the financial institution includes the following steps:
the method comprises the following steps: preprocessing data;
the data preprocessing comprises user data selection and data missing value processing, the data related to the sample in the embodiment of the invention comprises basic characteristics of the user such as age and gender, behavior preferences of the user such as transaction category distribution and financial product configuration situation, transaction detail data of the user, and tag information derived from the product holding situation of the user.
When user information data is acquired, data missing values need to be processed, data columns in a data table are observed, missing value columns are completed in a certain mode, for example, missing values of numerical characteristics are completed by column '0' values, missing values of non-numerical characteristics are completed by 'un' values, and for columns with particularly serious missing values, the field is directly deleted.
Step two: constructing sample characteristics;
extracting the transaction data of the user, such as transaction category distribution, financial product configuration condition and other behavior preferences as the behavior characteristics of the user by extracting the basic information of the user as the user characteristics, such as the age, the sex and other basic characteristics of the user; extracting basic information of the product as product characteristics; extracting transaction data of the product, such as transaction times, transaction modes and the like, as transaction characteristics of the product; and constructing sample characteristics based on the extracted basic user characteristics, behavior characteristics, product characteristics and product transaction characteristics.
Step three: sample training
The financial products related in the embodiment have a hierarchical relationship, for example, deposit products comprise regular periods and current periods, regular products comprise large-amount regular periods, 5-year regular periods, 3-year regular periods and the like, the product hierarchy effectively represents the relationship among different categories, a product recommendation model in the prior art only takes product attributes as product characteristics, and the hierarchical relationship is ignored. In this embodiment, three product levels are selected for schematic description (in this embodiment, the product level is not specifically limited, and may be set according to an actual application scenario).
Firstly, the calculation of a layering multi-label experience loss term is carried out on the selected three-layer product label, and a certain sample x belongs to the field ofX, which is labeled with a tertiary level L of { L } 1 ,l 2 ,l 3 Represents the three product levels that should be recommended for them, respectively, and the final model is composed of three-level classifiers f l () The method comprises the following steps of (1) forming a calculation formula of a hierarchical multi-label empirical loss term as follows:
wherein, X is a training sample set, X is a certain training sample therein, L is a multi-layer label of the training sample (namely determining the product level to which the training sample belongs), L is a certain layer sample label therein, f l () Classifiers corresponding to layer I labels, L emp Is a tag experience loss term.
In the embodiment, through acquiring the hierarchical multi-label experience loss, the machine learning model can perform fusion learning on label information of multiple hierarchies in the optimization process, and the generalization effect of the model is improved.
Secondly, positive and negative sample pairs are constructed for each layer of labels, if two training samples are on the l-th layer of labels and have the same label information, and both users hold a certain product, the two samples are positive sample pairs on the l-th layer of labels (corresponding to the training sample pairs which belong to the same product level and have the same sample labels being characterized as positive sample pairs), otherwise, the two samples are negative sample pairs (corresponding to the training sample pairs which do not belong to the same product level and/or have different sample labels being characterized as negative sample pairs).
In this embodiment, when the contrast loss value of the training sample pair is calculated, the calculation formula adopted is as follows:
wherein X and X' represent two training samples (corresponding to the above sample pair), X is the training sample set including all the training samples, f l () A classifier corresponding to the label of the l-th layer, a is different fromx other training samples, τ being a hyperparameter, L pair (x, x') is the contrast loss value of the two training samples.
Calculating a hierarchical multi-label contrast loss term by adopting the following formula:
wherein L is a product level, L is an L-th level sample label, X is a training sample set, X is a training sample in an X indication sample set, and P is l x To form a sample set of positive sample pairs with the training sample x,
is P l x One sample of the indicated set of samples;
for a set of samples that form negative sample pairs with sample x,
is composed of
One sample of the indicated set of samples;
represents sample x and sample
The value of the loss of contrast between,
representing a sample x and a sample
Loss of contrast between, L hmc The loss terms are compared for the hierarchical multi-label.
In this embodiment, all select the sample to each layer of label and carry out the study training, the model can make between the positive sample to be close to each other in the optimization process, keeps away from each other between the negative sample, through contrast study for the similar information between the model fully learns the positive sample and the discriminatory information between the negative sample, improves the model generalization effect.
After obtaining the hierarchical multi-label experience loss and the hierarchical multi-label comparison learning items, the final objective function of the hierarchical multi-label product recommendation model based on the comparison learning is as follows:
Loss＝L emp +L hmc ；
wherein Loss is an objective function, L emp For hierarchical multi-label empirical loss terms, L hmc The lossy terms are compared for hierarchical multi-labels.
Through optimization of the objective function, a hierarchical multi-label product recommendation model based on comparative learning can be obtained, and finally a classifier f of each product level is obtained l () And then, inputting the x to the trained classifier for the sample x to be tested, and obtaining the prediction result of each product level.
The embodiment of the invention provides a hierarchical multi-label multi-view product recommendation model based on contrast learning, each financial product has different levels of category attributes, so that a machine learning model can be optimized by fully utilizing label information of multiple levels, and the model can fully learn the level information of the product; the embodiment of the invention adopts a layered multi-label experience loss item, the model can fully learn the hierarchical information of the product, and adopts a layered multi-label comparison learning item, so that similar sample pairs are closer to each other in the label space of each level, and the sample pairs with large differences are farther away from each other in the label space of each level, thereby improving the generalization performance of the model.
The invention is described below in connection with an alternative embodiment.
Example two
The embodiment provides a product recommendation device of a financial institution, and each implementation module included in the product recommendation device of the financial institution corresponds to each implementation step in the first embodiment.
Fig. 4 is a schematic diagram of an alternative product recommendation apparatus of a financial institution according to an embodiment of the present invention, as shown in fig. 4, the product recommendation apparatus of the financial institution includes:
a receiving unit 41, configured to receive user data, where the user data includes: user characteristics, transaction detail data and transaction behaviors;
the input unit 42 is configured to input user data to a product recommendation model, and receive a recommendation result output by the product recommendation model, where training data of the product recommendation model is merged into label information of multiple product levels, each product level corresponds to a classifier, and the recommendation result includes: at least one quasi-recommended product and a product level to which each quasi-recommended product belongs;
and the display unit 43 is used for displaying the recommendation result on a system page.
The product recommendation device of the financial institution receives user data through the receiving unit 41, wherein the user data includes: user characteristics, transaction detail data and transaction behaviors are input into the product recommendation model through the input unit 42, and recommendation results output by the product recommendation model are received, wherein training data of the product recommendation model is integrated into label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation results include: the at least one quasi-recommended product and the product level to which each quasi-recommended product belongs are used for displaying the recommendation result on the system page through the display unit 43. In this embodiment, the label information of a plurality of product levels is merged into the product recommendation model, the attributes of different levels of the product are used as the basis for product recommendation, the diversity and accuracy of product recommendation are improved, the user requirements are fully met, the user experience is improved, and the technical problem that the recommendation effect is poor due to the fact that the hierarchical relationship between the products is ignored when the financial products are recommended by the recommendation model in the prior art is solved.
Optionally, the product recommendation device of the financial institution further comprises: a first obtaining subunit, configured to obtain a plurality of training samples, where each training sample at least includes: user characteristics, historical transaction behaviors and historical transaction detail data; the first construction subunit is used for constructing the sample characteristics of each training sample; the first determining subunit is used for determining a layered multi-label experience loss term and a layered multi-label comparison loss term corresponding to each training sample based on the sample characteristics; the second determining subunit is used for determining a model objective function based on the hierarchical multi-label empirical loss term and the hierarchical multi-label comparative loss term; and the first adjusting subunit is used for adjusting the model objective function to obtain the classifier of each product level and determine the product recommendation model.
Optionally, the first obtaining subunit includes: the first analysis module is used for analyzing data columns of the data table in each training sample; the first completion module is used for performing completion processing on missing values in the data columns; or, the first deleting module is configured to delete the data columns in which the total number of the missing items to which the missing values belong is greater than a preset number threshold.
Optionally, the first building subunit comprises: the first extraction module is used for extracting basic information related to each user in each training sample to obtain user characteristics of the user; the second extraction module is used for extracting behavior preference information related to each user in each training sample to obtain behavior characteristics; the third extraction module is used for extracting basic information of related transaction products in each training sample to obtain product characteristics; the fourth extraction module is used for extracting the transaction data of the associated transaction products in each training sample to obtain the product transaction characteristics; and the first construction module is used for constructing the sample characteristics of each training sample based on the user characteristics, the behavior characteristics, the product characteristics and the product transaction characteristics.
Optionally, the first determining subunit includes: a first determination module for determining a plurality of sample labels of the training sample based on the sample features; the first configuration module is used for configuring a plurality of product levels corresponding to the transaction products of the training samples and configuring a corresponding classifier for each product level; the second determining module is used for determining the product level to which the training sample belongs through the classifier based on the plurality of sample labels; the first establishing module is used for establishing a sample label layering model based on a sample set to which the training sample belongs, a plurality of sample labels, a product level to which the training sample belongs and a classifier corresponding to the product level, and determining a layering multi-label experience loss term of the training sample by the sample label layering model.
Optionally, the hierarchical multi-label experience loss item is used for characterizing hierarchical information of a product hierarchy to which the transaction product belongs and hierarchical association relations.
Optionally, the first determining subunit further includes: the second construction module is used for pairing all the training samples to construct a plurality of training sample pairs; the first query module is used for querying the sample label corresponding to each training sample in the training sample pair and the product level to which the sample label belongs; the first characterization module is used for characterizing training sample pairs which belong to the same product level and have the same sample label as a positive sample pair; the second characterization module is used for characterizing training sample pairs which do not belong to the same product level and/or have different sample labels as negative sample pairs; the first output module is used for outputting the generalization distance between the two training samples related to the positive sample pair in the sample set to be close and outputting the generalization distance between the two training samples related to the negative sample pair to be far by adopting a preset comparison learning strategy; the first learning module is used for controlling the product recommendation model to learn similar information between the training samples in the positive sample pair, learning distinguishing information between the training samples in the negative sample pair, and determining the hierarchical multi-label contrast loss items of the training samples through the product recommendation model.
The product recommendation device of the financial institution may further include a processor and a memory, where the receiving unit 41, the input unit 42, the display unit 43, and the like are stored in the memory as program units, and the processor executes the program units stored in the memory to implement corresponding functions.
The processor comprises a kernel, and the kernel calls the corresponding program unit from the memory. The kernel can be set to be one or more, a financial product recommendation model is built based on the hierarchical relation among the products by adjusting kernel parameters, and the financial products are recommended for the user.
The memory may include volatile memory in a computer readable medium, random Access Memory (RAM) and/or nonvolatile memory such as Read Only Memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.
According to another aspect of the embodiments of the present invention, there is also provided a computer-readable storage medium including a stored computer program, wherein when the computer program runs, the apparatus on which the computer-readable storage medium is located is controlled to execute any one of the above methods for recommending a product by a financial institution.
According to another aspect of embodiments of the present invention, there is also provided an electronic device including one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the product recommendation method for a financial institution of any one of the above.
The present application further provides a computer program product adapted to perform a program for initializing the following method steps when executed on a data processing device: receiving user data, wherein the user data comprises: user characteristics, transaction detail data and transaction behaviors; inputting user data into a product recommendation model, and receiving a recommendation result output by the product recommendation model, wherein training data of the product recommendation model is integrated into label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and a product level to which each quasi-recommended product belongs; and displaying the recommendation result on a system page.
Fig. 5 is a block diagram of a hardware configuration of an electronic device (or a mobile device) of a product recommendation method of a financial institution according to an embodiment of the present invention. As shown in fig. 5, the electronic device may include one or more (shown as 502a, 502b, \8230;, 502 n) processors 502 (processor 502 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), memory 504 for storing data. Besides, the method can also comprise the following steps: a display, an input/output interface (I/O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a keyboard, a power supply, and/or a camera. It will be understood by those skilled in the art that the structure shown in fig. 5 is only an illustration and is not intended to limit the structure of the electronic device. For example, the electronic device may also include more or fewer components than shown in FIG. 5, or have a different configuration than shown in FIG. 5.
The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages and disadvantages of the embodiments.
In the above embodiments of the present invention, the description of each embodiment has its own emphasis, and reference may be made to the related description of other embodiments for parts that are not described in detail in a certain embodiment.
In the embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other manners. The above-described embodiments of the apparatus are merely illustrative, and for example, a division of a unit may be a division of a logic function, and an actual implementation may have another division, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or may not be executed. In addition, the shown or discussed coupling or direct coupling or communication connection between each other may be an indirect coupling or communication connection through some interfaces, units or modules, and may be electrical or in other forms.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.
Claims (10)
1. A method for recommending products to a financial institution, comprising:
receiving user data, wherein the user data comprises: user characteristics, transaction detail data and transaction behaviors;
inputting the user data into a product recommendation model, and receiving a recommendation result output by the product recommendation model, wherein training data of the product recommendation model is integrated with label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and the product level to which each quasi-recommended product belongs;
and displaying the recommendation result on a system page.
2. The method of claim 1, when building the product recommendation model, comprising:
obtaining a plurality of training samples, wherein each training sample at least comprises: user characteristics, historical transaction behaviors, and historical transaction detail data;
constructing sample characteristics of each training sample;
determining a hierarchical multi-label experience loss term and a hierarchical multi-label comparison loss term corresponding to each training sample based on the sample characteristics;
determining a model objective function based on the hierarchical multi-label empirical loss term and the hierarchical multi-label comparison loss term;
and adjusting the model objective function to obtain a classifier of each product level, and determining the product recommendation model.
3. The method of claim 2, further comprising, after obtaining the plurality of training samples:
analyzing data columns of a data table in each training sample;
completing missing values in the data columns; alternatively, the first and second electrodes may be,
and deleting the data columns with the total number of the missing items of which the missing values belong to being larger than a preset number threshold.
4. The method of claim 2, wherein the step of constructing the sample features for each of the training samples comprises:
extracting basic information related to each user in each training sample to obtain user characteristics of the user;
extracting behavior preference information associated with each user in each training sample to obtain behavior characteristics;
extracting basic information of related transaction products in each training sample to obtain product characteristics;
extracting transaction data related to the transaction product in each training sample to obtain product transaction characteristics;
constructing a sample feature for each of the training samples based on the user feature, the behavior feature, the product feature, and the product transaction feature.
5. The method of claim 2, wherein the step of determining a hierarchical multi-label empirical loss term and a hierarchical multi-label comparative loss term for each of the training samples based on the sample features comprises:
determining a plurality of sample labels for the training sample based on the sample features;
configuring a plurality of the product levels corresponding to trading products of the training sample, and configuring the corresponding classifier for each product level;
determining, by the classifier, a product hierarchy to which the training sample belongs based on the plurality of sample labels;
establishing a sample label hierarchical model based on the sample set to which the training sample belongs, the plurality of sample labels, the product level to which the training sample belongs and the classifier corresponding to the product level, and determining a hierarchical multi-label experience loss term of the training sample by the sample label hierarchical model.
6. The method according to any one of claims 2 to 5, wherein the hierarchical multi-label experience loss term is used for representing hierarchical information of a product hierarchy to which a transaction product belongs and hierarchical association relations.
7. The method of claim 2, wherein the step of determining a hierarchical multi-label empirical loss term and a hierarchical multi-label contrast loss term corresponding to each of the training samples based on the sample features further comprises:
pairing all the training samples to construct a plurality of training sample pairs;
inquiring a sample label corresponding to each training sample in the training sample pair and a product level to which the sample label belongs;
characterizing the training sample pairs which belong to the same product level and have the same sample label as a positive sample pair;
characterizing the training sample pairs that do not belong to the same product hierarchy and/or have different sample labels as negative sample pairs;
adopting a preset comparison learning strategy, outputting the generalization distance between the two training samples related to the positive sample pair in the sample set to be close, and outputting the generalization distance between the two training samples related to the negative sample pair to be far;
and controlling the product recommendation model to learn similar information between the training samples in the positive sample pair, learning distinguishing information between the training samples in the negative sample pair, and determining a hierarchical multi-label contrast loss item of the training samples by the product recommendation model.
8. A product recommendation device for a financial institution, comprising:
a receiving unit, configured to receive user data, where the user data includes: user characteristics, transaction detail data and transaction behaviors;
the input unit is used for inputting the user data into a product recommendation model and receiving a recommendation result output by the product recommendation model, wherein the training data of the product recommendation model is fused with label information of a plurality of product levels, each product level corresponds to a classifier, and the recommendation result comprises: at least one quasi-recommended product and the product level to which each quasi-recommended product belongs;
and the display unit is used for displaying the recommendation result on a system page.
9. A computer-readable storage medium, comprising a stored computer program, wherein when the computer program is executed, the computer-readable storage medium is controlled to execute a product recommendation method of a financial institution of any one of claims 1 to 7.
10. An electronic device comprising one or more processors and memory storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of product recommendation for a financial institution of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211303505.6A CN115578165A (en) | 2022-10-24 | 2022-10-24 | Product recommendation method and device for financial institution, electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211303505.6A CN115578165A (en) | 2022-10-24 | 2022-10-24 | Product recommendation method and device for financial institution, electronic device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115578165A true CN115578165A (en) | 2023-01-06 |
Family
ID=84587813
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211303505.6A Pending CN115578165A (en) | 2022-10-24 | 2022-10-24 | Product recommendation method and device for financial institution, electronic device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115578165A (en) |
-
2022
- 2022-10-24 CN CN202211303505.6A patent/CN115578165A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070391B (en) | Data processing method and device, computer readable medium and electronic equipment | |
| CN109783730A (en) | Products Show method, apparatus, computer equipment and storage medium | |
| CN110163723A (en) | Recommended method, device, computer equipment and storage medium based on product feature | |
| CN110135942A (en) | Products Show method, apparatus and computer readable storage medium | |
| CN105744005A (en) | Client positioning and analyzing method and server | |
| CN113034246A (en) | Financial product recommendation method and device | |
| CN116468460B (en) | Consumer finance customer image recognition system and method based on artificial intelligence | |
| CN114663198A (en) | Product recommendation method, device and equipment based on user portrait and storage medium | |
| CN110852785A (en) | User grading method, device and computer readable storage medium | |
| CN111523315B (en) | Data processing method, text recognition device and computer equipment | |
| CN111680213B (en) | Information recommendation method, data processing method and device | |
| Nanayakkara et al. | A survey of finding trends in data mining techniques for social media analysis | |
| CN112395499B (en) | Information recommendation method and device, electronic equipment and storage medium | |
| CN111523914A (en) | User satisfaction evaluation method, device and system and data display platform | |
| CN113032676A (en) | Recommendation method and system based on micro-feedback | |
| CN110209944B (en) | Stock analyst recommendation method and device, computer equipment and storage medium | |
| CN116467525A (en) | Recommendation method, device, equipment and storage medium of business products | |
| CN113837843B (en) | Product recommendation method and device, medium and electronic equipment | |
| CN112015970A (en) | Product recommendation method, related equipment and computer storage medium | |
| CN115578165A (en) | Product recommendation method and device for financial institution, electronic device and storage medium | |
| CN115049456A (en) | Recommendation method and device for financial product combination and electronic equipment | |
| CN116764236A (en) | Game prop recommending method, game prop recommending device, computer equipment and storage medium | |
| CN114529399A (en) | User data processing method, device, computer equipment and storage medium | |
| CN113934612A (en) | User portrait updating method and device, storage medium and electronic equipment | |
| CN113191570A (en) | Fund planning recommendation method, device and equipment based on deep learning |
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 |