CN116362895A - Financial product recommendation method, device and storage medium - Google Patents

Financial product recommendation method, device and storage medium Download PDF

Info

Publication number
CN116362895A
CN116362895A CN202310539421.0A CN202310539421A CN116362895A CN 116362895 A CN116362895 A CN 116362895A CN 202310539421 A CN202310539421 A CN 202310539421A CN 116362895 A CN116362895 A CN 116362895A
Authority
CN
China
Prior art keywords
financial
tags
account
labels
random
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
Application number
CN202310539421.0A
Other languages
Chinese (zh)
Inventor
黄景超
曾炜
陈凌潇
容达锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Industrial and Commercial Bank of China Ltd ICBC
Original Assignee
Industrial and Commercial Bank of China Ltd ICBC
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Industrial and Commercial Bank of China Ltd ICBC filed Critical Industrial and Commercial Bank of China Ltd ICBC
Priority to CN202310539421.0A priority Critical patent/CN116362895A/en
Publication of CN116362895A publication Critical patent/CN116362895A/en
Pending legal-status Critical Current

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
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/06—Asset management; Financial planning or analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Development Economics (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Accounting & Taxation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Finance (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Economics (AREA)
  • General Business, Economics & Management (AREA)
  • Operations Research (AREA)
  • Computational Linguistics (AREA)
  • Technology Law (AREA)
  • Strategic Management (AREA)
  • Marketing (AREA)
  • Medical Informatics (AREA)
  • Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)

Abstract

The application discloses a financial product recommendation method, a financial product recommendation device and a storage medium. Relates to the field of financial science and technology, and the method comprises the following steps: acquiring first account feature data of a target account; based on the first account characteristic data, a random forest financial management prediction model is adopted to obtain financial management labels respectively output by N random trees and label scores corresponding to the financial management labels respectively output by the N random trees; according to the financial tags respectively output by the N random trees and the tag scores corresponding to the financial tags respectively output by the N random trees, obtaining the financial tag scores respectively corresponding to the K financial tags corresponding to the target account; and determining a financial product recommendation result of the target account based on the financial label scores corresponding to the K financial labels respectively. According to the method and the device for recommending financial products, the problems that the data size required to be used is large and recommendation is inaccurate in the method for recommending financial products through collaborative filtering recommendation strategies and knowledge-graph modes in the related technologies are solved.

Description

Financial product recommendation method, device and storage medium
Technical Field
The application relates to the field of financial science and technology, in particular to a financial product recommendation method, a financial product recommendation device and a storage medium.
Background
The financial product is taken as an important component of financial institution products, and a set of quick, accurate and effective recommendation system is provided, so that the service quality of the financial institution is greatly improved. In the prior art, rules of purchasing financial products by users are deeply mined through collaborative filtering recommendation strategies and knowledge maps, so that the financial products to be recommended are recommended to the users, and sales of the financial products are improved. However, collaborative filtering recommendation strategies require large amounts of data to analyze, and if the data is not sufficiently accurate, the recommendation accuracy is greatly affected. The method for recommending strategies and knowledge maps by collaborative filtering generally needs a large amount of data, the product sales field in the financial institutions is smaller, particularly the available data amount is limited, and other financial institutions related sales information is relatively closed, so that the acquisition cost is higher, the coupling degree of the method for training a large amount of data is lower, the network cannot be fully utilized, the network for training a large amount of data is often high in parameter or calculation amount, the number of data to be processed cannot be flexibly increased or decreased, and a large amount of pressure is caused on hardware resources. Knowledge maps also depend largely on the accessibility of data, and because of the huge amount of data far exceeding human processing capacity, numerous relationships hidden in the data cannot be found by human power.
Aiming at the problems of large data volume to be used and inaccurate recommendation in the method for recommending financial products by collaborative filtering recommendation strategies and knowledge graph modes in the related technologies, no effective solution is proposed at present.
Disclosure of Invention
The main purpose of the application is to provide a financial product recommendation method, a financial product recommendation device and a storage medium, so as to solve the problems of large data size and inaccurate recommendation in the related technology of the related technology, wherein the financial product recommendation method is performed through collaborative filtering recommendation strategies and knowledge graph modes.
In order to achieve the above object, according to one aspect of the present application, a financial product recommendation method is provided. The method comprises the following steps: acquiring first account feature data of a target account; based on the first account characteristic data, a random forest financial management prediction model is adopted to obtain financial management labels respectively output by N random trees and label scores corresponding to the financial management labels respectively output by the N random trees, wherein the random forest financial management prediction model is based on account characteristic data of M accounts, historical financial management behaviors of the M accounts are respectively obtained through training of the financial management labels, M, N is an integer greater than or equal to 2, and the random forest financial management prediction model comprises the N random trees; obtaining financial label scores corresponding to K financial labels corresponding to the target account according to the financial labels respectively output by the N random trees and label scores corresponding to the financial labels respectively output by the N random trees, wherein K is an integer greater than or equal to 2; and determining the financial product recommendation result of the target account based on the financial label scores corresponding to the K financial labels respectively.
In order to achieve the above object, according to another aspect of the present application, there is provided a financial product recommendation device. The device comprises: the first acquisition module is used for acquiring first account characteristic data of the target account; the prediction module is used for obtaining financial tags respectively output by N random trees and tag scores corresponding to the financial tags respectively output by the N random trees based on the first account characteristic data, wherein the random forest financial prediction model is obtained by training the financial tags respectively corresponding to the historical financial behaviors of M accounts based on the account characteristic data of the M accounts, M, N is an integer greater than or equal to 2, and the random forest financial prediction model comprises the N random trees; the second acquisition module is used for acquiring financial tag scores corresponding to K financial tags corresponding to the target account according to the financial tags respectively output by the N random trees and the tag scores corresponding to the financial tags respectively output by the N random trees, wherein K is an integer greater than or equal to 2; and the determining module is used for determining the financial product recommendation result of the target account based on the financial label scores respectively corresponding to the K financial labels.
In order to achieve the above object, according to another aspect of the present application, there is also provided a non-volatile storage medium storing a plurality of instructions adapted to be loaded and executed by a processor to any one of the financial product recommendation methods described above.
Through the application, the following steps are adopted: acquiring first account feature data of a target account; based on the first account characteristic data, a random forest financial management prediction model is adopted to obtain financial management labels respectively output by N random trees and label scores corresponding to the financial management labels respectively output by the N random trees, wherein the random forest financial management prediction model is based on account characteristic data of M accounts, historical financial management behaviors of the M accounts are respectively obtained through training of the financial management labels, M, N is an integer greater than or equal to 2, and the random forest financial management prediction model comprises the N random trees; obtaining financial label scores corresponding to K financial labels corresponding to the target account according to the financial labels respectively output by the N random trees and label scores corresponding to the financial labels respectively output by the N random trees, wherein K is an integer greater than or equal to 2; based on the financial tag scores corresponding to the K financial tags respectively, the financial product recommendation result of the target account is determined, the purposes of constructing a random forest financial prediction model to score the financial tags and accurately recommending the financial products according to the tag scoring result are achieved, and the problems that the data amount needed to be used is large and recommendation is inaccurate in a method for recommending the financial products through collaborative filtering recommendation strategies and knowledge graph modes in related technologies are solved. And the effects of reducing the training data quantity of the financial product recommendation model and improving the recommendation accuracy are achieved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application, illustrate and explain the application and are not to be construed as limiting the application. In the drawings:
FIG. 1 is a flow chart of a financial product recommendation method provided in accordance with an embodiment of the present application; and
FIG. 2 is a schematic diagram of an alternative financial product recommendation system according to an embodiment of the present application;
FIG. 3 is a schematic diagram of an alternative random forest financial prediction model training process according to an embodiment of the present application;
FIG. 4 is a schematic diagram of an alternative financial product recommendation process according to an embodiment of the present application;
FIG. 5 is a schematic diagram of a financial product recommendation device provided according to an embodiment of the present application;
fig. 6 is a schematic diagram of an electronic device provided according to an embodiment of the present application.
Detailed Description
It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate in order to describe the embodiments of the present application 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, related information (including, but not limited to, user equipment information, user personal information, etc.) and data (including, but not limited to, data for presentation, analyzed data, etc.) related to the present application 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 institution, before acquiring the relevant information, the system needs to send an acquisition request to the aforesaid user institution via the interface, and acquire the relevant information after receiving the consent information fed back by the aforesaid user or institution.
In the following description of the preferred implementation steps, fig. 1 is a flowchart of a financial product recommendation method according to an embodiment of the present application, as shown in fig. 1, and the method includes the following steps:
step S101, obtaining first account feature data of a target account.
Optionally, the account feature data (i.e. the first feature data) corresponding to the target account is extracted based on information such as personal basic information of the client, financial risk assessment report, personal financial behavior habit, personal credit report, financial product search preference and the like.
Step S102, based on the first account feature data, a random forest financial management prediction model is adopted to obtain financial tags respectively output by N random trees and tag scores corresponding to the financial tags respectively output by the N random trees, wherein the random forest financial management prediction model is obtained by training the financial tags respectively corresponding to the historical financial management behaviors of M accounts based on the account feature data of the M accounts, M, N is an integer greater than or equal to 2, and the random forest financial management prediction model comprises the N random trees.
Optionally, the financial tag is used to indicate the attribute of the financial product, such as various attributes of the financial product, such as risk high-low attribute, type (deposit, fund, stock, foreign exchange, etc.), term (flexible, 1 month, 3 months, 1 year, etc.), income type, point of purchase, transaction currency type, issuer and manager, etc., and in fact, the financial product tag is far more than these, possibly thousands of tags. The label score corresponding to the financial label is used for indicating the quantity ratio of the labels, namely the proportion of the quantity of each type of label to the total quantity of the labels. If there are 10 tags, there are 5 tag financial product recommendation methods, devices and storage media, and 5 tags Y, then the tag scores of the financial product recommendation methods, devices and storage media and Y are all 0.5.
In the related art, a large amount of data is required to be used for analysis in the mode of collaborative filtering recommendation strategy, knowledge graph and the like for financial product recommendation, through the mode, based on the first account characteristics corresponding to the target account, a pre-trained random forest financial prediction model is adopted, N random trees can be output to respectively conduct label prediction, and financial label scores of each financial label corresponding to the target account are comprehensively obtained according to the financial labels respectively output by the N random trees and the corresponding label scores. And moreover, a random forest financial prediction model is obtained based on random forest mode training, the data volume required in the model training stage is far smaller than the data volume required by financial product recommendation through collaborative filtering recommendation strategies, knowledge maps and other modes, and the defects that a large amount of data are required to be used for analysis and the data are bulky in the prior art are effectively overcome.
In an optional embodiment, before the first account feature data based on the target account uses a random forest financial management prediction model to obtain financial tags respectively output by N random trees and tag scores corresponding to the financial tags respectively output by the N random trees, the method further includes: acquiring account characteristic data of the M accounts and financial tags corresponding to historical financial behaviors of the M accounts respectively; based on the account characteristic data of the M accounts and financial labels corresponding to the historical financial behaviors of the M accounts, carrying out L rounds of node division according to the financial labels to generate the N random trees, wherein L is an integer greater than or equal to 2; and obtaining the random forest financial prediction model according to the N random trees.
Optionally, the training process of the random forest financial management prediction model specifically includes: based on account feature data of M accounts and financial labels corresponding to historical financial behaviors of the M accounts, starting from a root node, performing multi-round node division according to the financial labels, and correspondingly outputting N random trees under the condition that the number of the financial labels and the number of the account feature data included in each leaf node obtained by node division of the current round are within a specified number, wherein the N random trees are combined to obtain a random forest financial prediction model.
In an optional embodiment, the generating the N random trees based on the account feature data of the M accounts and financial tags corresponding to historical financial behaviors of the M accounts respectively includes: for any current round of node division in the L rounds of node division, based on account feature data of the M accounts and financial tags respectively corresponding to historical financial behaviors of the M accounts, performing the current round of node division according to the financial tags until the number of the financial tags and the number of the account feature data included in each leaf node obtained by the current round of node division are in a corresponding preset number range: according to financial labels respectively corresponding to the historical financial behaviors of the M accounts, account feature data corresponding to the dividing nodes of the previous round of the current round are divided into two leaf nodes, wherein the account feature data corresponding to the dividing nodes of the previous round are account feature data of the M accounts when the dividing nodes of the previous round are root nodes; determining the types of financial tags respectively included in the two leaf nodes and the number of each financial tag; according to the quantity of each financial label, sorting the financial labels respectively included in the two leaf nodes to obtain a first sorting result; and re-dividing the account feature data corresponding to the two leaf nodes according to the first sorting result to obtain new two leaf nodes and a second sorting result until the first sorting result is the same as the second sorting result.
In an alternative embodiment, the method further comprises: and repeatedly executing the operations of re-node division of account feature data respectively included by the two leaf nodes according to the first sorting result to obtain new two leaf nodes and the operation of the second sorting result until the second sorting result is the same as the first sorting result by taking the new two leaf nodes as the two leaf nodes and taking the second sorting result as the first sorting result under the condition that the second sorting result is different from the first sorting result.
Optionally, the single random tree starts from a root node, the root node includes account feature data and financial tags of M accounts, and the financial tags are managed to perform multi-round node division to generate a random forest financial prediction model, and the method specifically includes the following steps:
step S1, splitting is started from a root node, client sample data (comprising account feature data corresponding to M accounts and financial labels) of a current splitting node are randomly initialized (divided) according to the financial labels, the client sample data are randomly divided into left and right child nodes, and the financial labels of the left and right child nodes are ordered according to the number of the labels, so that a first ordering result is obtained.
Step S2, the client sample data is divided again according to the first sorting result, and the dividing basis is as follows: the M accounts respectively correspond to the quantity of the financial tags in the sorting of the financial tags, and the quantity is more than the quantity. And re-dividing each client sample data in the mode to obtain a new round of divided left and right child nodes and a corresponding second sorting result.
And step S3, repeating the step S2 to carry out node division again on each piece of customer sample data of the previous round. And repeating the financial tag sorting and the left and right child node re-dividing steps until the front and back dividing results are the same, and completing the current node dividing.
And S4, after the current node is divided, repeating the steps S11 to S13 for each divided leaf node. Termination conditions for random tree training: the partitioning proceeds until each leaf node contains a financial tag and the number of customer samples within a specified number.
And S5, repeating the steps S2 to S4, training to form a plurality of random trees, and integrating a random forest financial prediction model.
In an optional embodiment, the obtaining, by using a random forest financial prediction model, the financial tags output by the N random trees respectively and tag scores corresponding to the financial tags output by the N random trees respectively based on the first account feature data of the target account includes: based on the first account characteristic data of the target account, the random forest financial management prediction model is adopted to obtain financial management labels output by at least one leaf node corresponding to the N random trees respectively and label scores corresponding to the financial management labels output by the at least one leaf node; and obtaining the financial labels respectively output by the N random trees and the label scores corresponding to the financial labels respectively output by the N random trees according to the financial labels respectively output by at least one leaf node of the N random trees and the first label scores corresponding to the financial labels output by the at least one leaf node.
Optionally, the target account finally reaches the leaf node through the prediction path of the random tree to obtain the label score of the corresponding financial label of the leaf node, and the label score corresponding to the financial label output by each random tree is obtained through the label score summary of the financial label of the leaf node. The label score corresponding to each financial label which is obtained by independent prediction of each random tree is obtained through random forest integration, the financial label score of each financial label corresponding to a target account is obtained, and the prediction error of a single tree is reduced. The financial label and the label score obtained by prediction of each random tree are possibly different, so that the financial label score obtained by integration of random forests is high in precision.
Step S103, obtaining the financial label scores corresponding to the K financial labels corresponding to the target account according to the financial labels respectively output by the N random trees and the label scores corresponding to the financial labels respectively output by the N random trees, wherein K is an integer greater than or equal to 2.
It should be noted that, the random forest financial management prediction model in the application comprises a plurality of random trees, and the label scores obtained by independent prediction of each random tree are integrated through the random forest to obtain integrated financial management label scores, so that the prediction error of a single tree is reduced. Compared with a financial recommendation system realized by using a single decision tree, the recommendation effect is good, recommendation errors and risks are effectively reduced, for example, the possibility that high-risk financial products are incorrectly recommended to clients with low risk bearing capacity is reduced, and unnecessary losses of the clients are reduced.
Step S104, determining the financial product recommendation result of the target account based on the financial label scores corresponding to the K financial labels respectively.
In an optional embodiment, the determining the financial product recommendation result of the target account based on the financial tag scores corresponding to the K financial tags respectively includes: updating the financial tag scores corresponding to the K financial tags respectively based on a preset score threshold value to obtain new financial tag scores corresponding to the K financial tags respectively; normalizing the new financial label scores corresponding to the K financial labels respectively to obtain the financial label probabilities corresponding to the K financial labels respectively; according to the probabilities of the financial tags corresponding to the K financial tags, sorting the K financial tags to obtain a third sorting result; and determining the financial product recommendation result of the target account according to the third sorting result.
Optionally, the result returned after random forest integration is a financial tag score related to a financial tag, and the new financial tag score of the client is obtained by setting a threshold value: setting a preset score threshold, keeping the score larger than the preset score threshold unchanged, and setting the score smaller than the preset score threshold to zero to obtain a new financial label score. And carrying out normalization operation on the new financial tag score to obtain the financial tag probability. The normalization operation here refers to calculating the proportion of each financial label score of the target account to the sum of all financial label scores, that is, the sum of all probabilities is one. And removing the financial tag with the score of zero in the obtained new financial tag score (zero setting operation), wherein the rest financial tags are financial prediction tags (one setting operation) of clients, and sorting the financial tags according to the probability of the financial tags to obtain a financial product recommendation result facing to the target account.
In an optional embodiment, the updating the financial tag scores corresponding to the K financial tags respectively based on the preset score threshold value, to obtain new financial tag scores corresponding to the K financial tags corresponding to the target account respectively includes: determining financial tags smaller than the preset score threshold in the financial tag scores corresponding to the K financial tags respectively; and replacing the financial label score corresponding to the financial label smaller than the preset score threshold value with a preset value to obtain new financial label scores corresponding to the K financial labels respectively.
Optionally, a preset score threshold is set, the score greater than the preset score threshold remains unchanged, and the score smaller than the preset score threshold is set to zero, so that a new financial label score is obtained. Therefore, interference of financial tags with lower financial tag scores on financial product recommendation results is eliminated.
In an optional embodiment, the sorting processing is performed on the K financial tags according to the probabilities of the financial tags corresponding to the K financial tags, to obtain a sorting result, where the sorting result includes: removing financial tags with probability smaller than a preset probability threshold from the K financial tags; and sorting other financial tags except for the financial tag with the probability smaller than the preset probability threshold value in the K financial tags to obtain the sorting result.
Through the method, the financial tags with lower probability (lower than a preset probability threshold) in the K financial tags are removed, only the financial tags with higher probability are reserved, and the financial tags with higher probability are sorted, so that the obtained sorting result is the recommendation sequence of the financial products.
In an optional embodiment, after determining the financial product recommendation result of the target account based on the financial tag scores corresponding to the K financial tags respectively, the method further includes: receiving a grading value of the financial product included in the financial product recommendation result aiming at the target account; and removing financial products with score values smaller than a preset score threshold value from the financial products included in the financial product recommendation result to obtain a new financial product recommendation result.
By the method, after the financial product recommendation result is pushed to the target account, the target account can score each financial product included in the financial product recommendation result, and the financial product with the excessively low score value can be removed from the financial product recommendation result to obtain a new financial product recommendation result and the new financial product recommendation result is pushed to the target account.
Through the steps S101 to S104, the purposes of constructing a random forest financial management prediction model to score financial management labels and accurately recommending financial management products according to label scoring results can be achieved, and the problems that in the related technology, the data amount needed to be used is large and recommendation is inaccurate in a method for recommending the financial management products through collaborative filtering recommendation strategies and knowledge graph modes in the related technology are solved. Thereby achieving the effects of reducing the training data quantity of the financial product recommendation model and improving the recommendation accuracy.
Based on the foregoing embodiments and optional embodiments, an optional implementation manner is provided in the present application, fig. 2 is a schematic structural diagram of an optional financial product recommendation system according to an embodiment of the present application, and as shown in fig. 2, the financial product recommendation system 200 includes a preprocessing module 202, a training module 204, a prediction module 206, and an iterative optimization module 208, where,
the preprocessing module 202 is configured to collect information such as customer personal basic information, financial risk assessment reports, personal financial behavior habits, personal credit reports, financial product search preferences of the M accounts from the database. And the acquired data sets are arranged into one-to-one corresponding customer information portraits and customer historical financial behavior sequences, the customer information portraits are used as account characteristic data of M accounts, and the customer historical financial behavior sequences are used as financial labels, so that subsequent model training is facilitated.
The training module 204 is used for training a plurality of random trees to form a random forest. The single random tree starts from a root node, the root node comprises account characteristic data and financial tags of M accounts, and the specific implementation process is as follows:
step S11, splitting is started from a root node, client sample data (comprising account feature data corresponding to M accounts and financial labels) of a current splitting node are randomly initialized (divided) according to the financial labels, the client sample data are randomly divided into left and right child nodes, and the financial labels of the left and right child nodes are ordered according to the number of the labels, so that a first ordering result is obtained.
Step S12, the client sample data is re-divided according to the first sorting result, wherein the dividing basis is as follows: the M accounts respectively correspond to the quantity of the financial tags in the sorting of the financial tags, and the quantity is more than the quantity. And re-dividing each client sample data in the mode to obtain a new round of divided left and right child nodes and a corresponding second sorting result.
Step S13, repeating step S12 to re-divide the nodes of each customer sample data of the previous round. And repeating the financial tag sorting and the left and right child node re-dividing steps until the front and back dividing results are the same, and completing the current node dividing.
Step S14, after the current node is divided, repeating the steps S11 to S13 for each divided leaf node. Termination conditions for random tree training: the partitioning proceeds until each leaf node contains a financial tag and the number of customer samples within a specified number.
And step S15, repeating the steps S12 to S14, training to form a plurality of random trees, and integrating a random forest financial prediction model.
The prediction module 206 is configured to input first account feature data corresponding to a target account of a financial product to be recommended into a random forest financial prediction model, predict each random tree of the random forest separately, and return a financial tag score corresponding to each financial tag after integrating the random forest, thereby obtaining a probability matrix and a financial tag matrix, and specifically includes the following steps:
and S21, the target account finally reaches the leaf node through the prediction path of the random tree to obtain the label score of the corresponding financial label of the leaf node, and the label scores of the financial labels of the leaf node are summarized to obtain the label scores corresponding to the financial labels respectively output by the random trees.
Step S22, the label scores corresponding to the financial labels obtained through independent prediction of the random trees are obtained through random forest integration, the financial label score of each financial label corresponding to the target account is obtained, and the prediction error of a single tree is reduced. The financial label and the label score obtained by prediction of each random tree are possibly different, so that the financial label score obtained by integration of random forests is high in precision.
Step S23, the result returned after random forest integration is a financial tag score related to the financial tag, and a new financial tag score of the client is obtained by setting a threshold value: setting a preset score threshold, keeping the score larger than the preset score threshold unchanged, and setting the score smaller than the preset score threshold to zero to obtain a new financial label score.
Step S24, obtaining the financial tag probability of the customer from the new financial tag score by the following method: and carrying out normalization operation on the new financial tag score to obtain the financial tag probability. The normalization operation here refers to calculating the proportion of each financial label score of the target account to the sum of all financial label scores, that is, the sum of all probabilities is one. The predicted financial label of the customer is obtained from the new financial label score by: and removing the financial label with zero score in the obtained new financial label score (zero setting operation), wherein the rest financial labels are financial prediction labels of clients (one setting operation).
Step S25, sorting the financial tags according to the probability of the financial tags of the target account, and displaying the financial tags on a financial product recommendation system.
The iterative optimization module 208 is configured to demonstrate and subsequently optimize the model, specifically using the steps of:
step S31, a recommended financial product display list is obtained from a financial product recommendation system, and a financial manager recommends and introduces the list as a basis for a user.
And step S32, sequentially adjusting the recommended products of the list according to the satisfaction degree of the user, manually removing if unsatisfactory products exist, and transmitting the final result back to the financial product recommendation system.
And step S33, acquiring the input data of the secondary financial prediction as characteristic input, and taking a final feedback list as a new financial tag list to continuously perform iterative optimization training on the model.
Based on the foregoing embodiments and optional embodiments, an optional implementation manner is provided in the present application, and fig. 3 is a schematic diagram of an optional random forest financial management prediction model training process according to an embodiment of the present application, as shown in fig. 3, assuming that financial tags in existing customer sample data respectively include: for convenience, the financial institution deposit is represented by a, the bond is represented by B, the insurance is represented by C, the fund is represented by D, the stock is represented by E, and the noble metal is represented by F. The labels in the above figures are explained as follows:
301 is the left child node financial tag ordering: the sorting result of the left child node in the diagram is that there are three financial tags a, two financial tags C, D, and one financial tag B, E, F. The number of financial tags is obtained by summing the same financial tag number in the customer sample data respectively scheduled to be divided to the left child node.
302 is customer sample data divided into left child nodes: the rule of division is a random and uniform division, where the customer sample data divided into the left child node is marked with a "box" for better differentiation.
303 is the total customer sample space for the current node: all client sample data under the current node are contained, and the client sample data contains account feature data and financial tags of M accounts. And if the root node is the root node, the root node contains all the customer sample data sets for training. The model training herein is to sort financial tags so that customer features are not labeled for visual convenience, but customer features are the criteria for making decisions when predicting recommendations.
Denoted at 304 is customer sample data divided into right child nodes: the rule of division is a random and uniform division, where the customer sample data divided to the right child node is marked with a "circle" for better differentiation.
Denoted at 305 is right child node financial tag ordering: the sorting result of the right child node in the diagram is that there are three financial tags a, two financial tags C, D, and one financial tag A, E, F. The number of financial tags is obtained by summing the same financial tag number in the customer sample data respectively scheduled to be divided to the right child node.
The random forest financial management prediction model training process is as follows:
in step S41, in 2-1 of fig. 3, the (root) node starts splitting, all the customer sample data are randomly and uniformly divided into two leaf nodes, namely, left and right child nodes, and the financial label number of the left child node is counted: a=3, c=d=2, b=e=f=1, counting the financial tags number of the right child node: b=3, c=d=2, a=e=f=1.
Step S42, sorting the financial labels, and re-dividing the nodes according to the sorting result of the round of financial labels, wherein the nodes are winning in a large number. Meaning that the financial tags are reclassified by taking the number of the financial tags as the ordering standard from the most financial tags. As shown in fig. 3 at 2-1, the left child node has more a than the right child node, the customer sample data related to a should BE re-divided into the left child node, the right child node has more B than the left child node, the customer sample data related to B should BE re-divided into the right child node, so that three ACs and one AF are re-divided into the left child node, three BDs and one BE are re-divided into the right child node, and one CF and one DE, because the sorting result of C, D, E, F financial tags in the present round of financial tag sorting is the same, thus keeping the original division result unchanged. The repartitioning results of this round are shown in fig. 3, 2-2.
Step S43, repeating the step S42, sorting the new round of financial labels, and continuing to divide again according to the sorting result of the financial labels. The results of 2-3 of fig. 3 were then obtained.
And step S44, repeating the financial tag sorting and re-dividing steps until the front and back dividing results are the same, and completing the current node division. As in 2-4 of fig. 3, the current node partitioning is complete. The computation separator divides the current node into left and right child nodes.
Step S45, repeating steps S41 to S44 for the left child node and the right child node obtained in step 2-4 of FIG. 3, and dividing termination conditions: the partitioning proceeds until each leaf node contains a financial tag and the number of customer samples within a specified number. Such a random tree is generated with the number of financial tags within a leaf node being the financial tag score (matrix) for that leaf node.
And S46, repeating the steps S41 to S45, generating a plurality of random trees, and integrating a random forest financial prediction model. The node change procedure pair of the four trees below in fig. 3 corresponds to node division procedures of 2-1,2-2,2-3,2-4, respectively.
Based on the foregoing embodiments and optional embodiments, an optional implementation manner is provided in the present application, and fig. 4 is a schematic diagram of an optional financial product recommendation process according to an embodiment of the present application, as shown in fig. 4, where a financial product recommendation flow is as follows:
And S51, the clients to be recommended (corresponding to the target accounts) respectively reach leaf nodes through all random trees in the random forest according to the corresponding first account characteristic data, gray arrows are the financial product prediction paths of each tree, and the label scores of financial labels of the leaf nodes where different trees are located are obtained. For example, tree 1 financing predicts the label score of ABCD, tree 2 financing predicts the label score of AEB, and tree 3 financing predicts the label score of AEF. Wherein the height of the histogram represents the score value size.
In step S52, the label scores obtained by individually predicting the financial products of each random tree are integrated through the random forest to obtain the integrated financial label scores corresponding to each financial label, so that the error of predicting the financial products of a single tree is reduced. For example, the integrated financial label score is shown in the AEBFDC financial label score histogram below fig. 4.
Step S53, obtaining new financial tag scores of the clients to be recommended by setting a threshold value: setting a threshold value, keeping the score larger than the threshold value unchanged, and setting the score smaller than the threshold value to zero to obtain a new financial label score. For fig. 4, it can be understood that a horizontal line (threshold) is drawn on the AEBFDC financial management label score bar graph, the score below the horizontal line is zeroed out, and the score above the horizontal line remains.
Step S54, obtaining financial tag probability of each financial tag corresponding to the customer to be recommended from the new financial tag score: and carrying out normalization operation on the new financial tag score to obtain the financial tag probability. The normalization operation here refers to calculating the proportion of each financial label score of the customer to be recommended to the sum of all financial label scores, that is, the sum of all probabilities is one. Obtaining financial tags (obtained by financial prediction) of clients to be recommended from the new financial tag scores: and removing the financial tag with the score of zero in the obtained new financial tag score (zero setting operation), wherein the rest financial tags are the financial tags predicted by the financial recommendation system (one setting operation).
And step S55, sorting financial prediction labels according to the financial label probability of the clients obtained in the step S54, and displaying the corresponding financial products on a financial product recommendation system.
It should be noted that, at least the following technical effects may be achieved in the embodiments of the present application: 1. the method has the advantages that the coupling degree of the small sample data is good, the information island situation can be effectively treated, and the situation that recommendation is completely deviated due to insufficient training is avoided. In the initial stage of online of the financial recommendation system, under the condition that the number of system users is relatively small, the effective samples acquired by the system are relatively small, and the coupling capacity of the system to the small samples enables the financial recommendation system to have higher accuracy in the early stage of online. 2. In the method, a plurality of random trees are generated through training, the label scores obtained through independent prediction of each random tree are integrated through random forests to obtain integrated financial label scores, and the prediction error of a single tree is reduced. Compared with the financial recommendation system realized by using a single decision tree, the financial recommendation system realized by using the method has good recommendation effect, effectively reduces recommendation errors and risks, for example, reduces the possibility that high-risk financial products are incorrectly recommended to clients with low risk bearing capacity, and reduces unnecessary loss of the clients. 3. The training effect on large sample data is also better, and when the sample size is changed greatly, the change amplitude of the recommended accuracy is small, and the prediction precision is stable and high. By using the financial system for realizing the product recommendation function, the influence of fluctuation of the quantity of financial clients on the accuracy is small, so that the financial recommendation system can keep higher accuracy no matter the quantity of the financial clients is increased or reduced, and the capability is particularly important especially under the condition that the quantity of active users is large. 4. The model training and prediction recommendation method and device are less in time consumption and higher in efficiency, and can bring better user experience. Specific benefits of the present application that are highly efficient include, but are not limited to, the following scenarios: when a new customer uses the system for the first time, the system can rapidly predict financial preferences of the customer according to the customer information and immediately recommend financial products matched with the customer; after the customer generates financial behaviors (purchase, conversion, redemption and the like), the customer can rapidly predict again according to the just-described customer behaviors and dynamically adjust the type and priority order of recommended financial products; on the premise of high training efficiency, the recommendation model can be iterated more frequently, and after a certain number of clients with financial behaviors reach, the system is triggered to retrain the recommendation model, so that the recommendation accuracy is improved.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.
The embodiment of the application also provides a financial product recommendation device, and it is to be noted that the financial product recommendation device of the embodiment of the application can be used for executing the financial product recommendation method provided by the embodiment of the application. The financial product recommendation device provided by the embodiment of the application is described below.
Fig. 5 is a schematic diagram of a financial product recommendation device according to an embodiment of the present application. As shown in fig. 5, the apparatus includes: a first acquisition module 500, a prediction module 502, a second acquisition module 504, a determination module 506, wherein,
the first obtaining module 500 is configured to obtain first account feature data of a target account;
the prediction module 502 is connected to the first obtaining module 500, and is configured to obtain, based on the first account feature data, financial tags output by N random trees respectively and tag scores corresponding to the financial tags output by the N random trees respectively, where the random forest financial prediction model is obtained by training the financial tags corresponding to the historical financial behaviors of M accounts respectively, where M, N is an integer greater than or equal to 2, and the random forest financial prediction model includes the N random trees;
The second obtaining module 504 is connected to the predicting module 502, and is configured to obtain, according to the financial tags respectively output by the N random trees and the tag scores corresponding to the financial tags respectively output by the N random trees, financial tag scores corresponding to the K financial tags corresponding to the target account respectively, where K is an integer greater than or equal to 2;
the determining module 506 is connected to the second obtaining module 504, and is configured to determine a financial product recommendation result of the target account based on the financial tag scores corresponding to the K financial tags.
According to the financial product recommendation device provided by the embodiment of the application, the first obtaining module 500 is configured to obtain the first account feature data of the target account; the prediction module 502 is connected to the first obtaining module 500, and is configured to obtain, based on the first account feature data, financial tags output by N random trees respectively and tag scores corresponding to the financial tags output by the N random trees respectively, where the random forest financial prediction model is obtained by training the financial tags corresponding to the historical financial behaviors of M accounts respectively, where M, N is an integer greater than or equal to 2, and the random forest financial prediction model includes the N random trees; the second obtaining module 504 is connected to the predicting module 502, and is configured to obtain, according to the financial tags respectively output by the N random trees and the tag scores corresponding to the financial tags respectively output by the N random trees, financial tag scores corresponding to the K financial tags corresponding to the target account respectively, where K is an integer greater than or equal to 2; the determining module 506 is connected to the second obtaining module 504, and is configured to determine a financial product recommendation result of the target account based on the financial tag scores corresponding to the K financial tags, so as to achieve the purpose of constructing a random forest financial prediction model to score the financial tag, and accurately recommending the financial product according to the tag scoring result, thereby solving the problems of large data volume to be used and inaccurate recommendation in the related art. And the effects of reducing the training data quantity of the financial product recommendation model and improving the recommendation accuracy are achieved.
It should be noted that each of the above modules may be implemented by software or hardware, for example, in the latter case, it may be implemented by: the above modules may be located in the same processor; alternatively, the various modules described above may be located in different processors in any combination.
It should be noted that, the first obtaining module 500, the predicting module 502, the second obtaining module 504, and the determining module 506 correspond to steps S101 to S104 in the embodiment, and the modules are the same as the examples and application scenarios implemented by the corresponding steps, but are not limited to the disclosure of the foregoing embodiment. It should be noted that the above modules may be run in a computer terminal as part of the apparatus.
It should be noted that, the optional or preferred implementation manner of this embodiment may be referred to the related description in the embodiment, and will not be repeated herein.
The financial product recommendation device comprises a processor and a memory, wherein the units and the like are stored in the memory as program units, and the processor executes the program units stored in the memory to realize corresponding functions.
The processor includes a kernel, and the kernel fetches the corresponding program unit from the memory. The kernel may be provided with one or more kernel parameters (for the purposes of this application).
The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip.
The embodiment of the application provides a nonvolatile storage medium, wherein a program is stored on the nonvolatile storage medium, and the program realizes the financial product recommendation method when being executed by a processor.
The embodiment of the application provides a processor, which is used for running a program, wherein the financial product recommendation method is executed when the program runs.
As shown in fig. 6, an embodiment of the present application provides an electronic device, where the electronic device 10 includes a processor, a memory, and a program stored on the memory and executable on the processor, and the processor implements the following steps when executing the program: acquiring first account feature data of a target account; based on the first account characteristic data, a random forest financial management prediction model is adopted to obtain financial management labels respectively output by N random trees and label scores corresponding to the financial management labels respectively output by the N random trees, wherein the random forest financial management prediction model is based on account characteristic data of M accounts, historical financial management behaviors of the M accounts are respectively obtained through training of the financial management labels, M, N is an integer greater than or equal to 2, and the random forest financial management prediction model comprises the N random trees; obtaining financial label scores corresponding to K financial labels corresponding to the target account according to the financial labels respectively output by the N random trees and label scores corresponding to the financial labels respectively output by the N random trees, wherein K is an integer greater than or equal to 2; and determining the financial product recommendation result of the target account based on the financial label scores corresponding to the K financial labels respectively. The device herein may be a server, PC, PAD, cell phone, etc.
The present application also provides a computer program product adapted to perform, when executed on a data processing device, a program initialized with the method steps of: acquiring first account feature data of a target account; based on the first account characteristic data, a random forest financial management prediction model is adopted to obtain financial management labels respectively output by N random trees and label scores corresponding to the financial management labels respectively output by the N random trees, wherein the random forest financial management prediction model is based on account characteristic data of M accounts, historical financial management behaviors of the M accounts are respectively obtained through training of the financial management labels, M, N is an integer greater than or equal to 2, and the random forest financial management prediction model comprises the N random trees; obtaining financial label scores corresponding to K financial labels corresponding to the target account according to the financial labels respectively output by the N random trees and label scores corresponding to the financial labels respectively output by the N random trees, wherein K is an integer greater than or equal to 2; and determining the financial product recommendation result of the target account based on the financial label scores corresponding to the K financial labels respectively.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: acquiring account characteristic data of the M accounts and financial tags corresponding to historical financial behaviors of the M accounts respectively; based on the account characteristic data of the M accounts and financial labels corresponding to the historical financial behaviors of the M accounts, carrying out L rounds of node division according to the financial labels to generate the N random trees, wherein L is an integer greater than or equal to 2; and obtaining the random forest financial prediction model according to the N random trees.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: for any current round of node division in the L rounds of node division, based on account feature data of the M accounts and financial tags respectively corresponding to historical financial behaviors of the M accounts, performing the current round of node division according to the financial tags until the number of the financial tags and the number of the account feature data included in each leaf node obtained by the current round of node division are in a corresponding preset number range: according to financial labels respectively corresponding to the historical financial behaviors of the M accounts, account feature data corresponding to the dividing nodes of the previous round of the current round are divided into two leaf nodes, wherein the account feature data corresponding to the dividing nodes of the previous round are account feature data of the M accounts when the dividing nodes of the previous round are root nodes; determining the types of financial tags respectively included in the two leaf nodes and the number of each financial tag; according to the quantity of each financial label, sorting the financial labels respectively included in the two leaf nodes to obtain a first sorting result; and re-dividing the account feature data corresponding to the two leaf nodes according to the first sorting result to obtain new two leaf nodes and a second sorting result until the first sorting result is the same as the second sorting result.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: and repeatedly executing the operations of re-node division of account feature data respectively included by the two leaf nodes according to the first sorting result to obtain new two leaf nodes and the operation of the second sorting result until the second sorting result is the same as the first sorting result by taking the new two leaf nodes as the two leaf nodes and taking the second sorting result as the first sorting result under the condition that the second sorting result is different from the first sorting result.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: based on the first account characteristic data of the target account, the random forest financial management prediction model is adopted to obtain financial management labels output by at least one leaf node corresponding to the N random trees respectively and label scores corresponding to the financial management labels output by the at least one leaf node; and obtaining the financial labels respectively output by the N random trees and the label scores corresponding to the financial labels respectively output by the N random trees according to the financial labels respectively output by at least one leaf node of the N random trees and the first label scores corresponding to the financial labels output by the at least one leaf node.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: updating the financial tag scores corresponding to the K financial tags respectively based on a preset score threshold value to obtain new financial tag scores corresponding to the K financial tags respectively; normalizing the new financial label scores corresponding to the K financial labels respectively to obtain the financial label probabilities corresponding to the K financial labels respectively; according to the probabilities of the financial tags corresponding to the K financial tags, sorting the K financial tags to obtain a third sorting result; and determining the financial product recommendation result of the target account according to the third sorting result.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: determining financial tags smaller than the preset score threshold in the financial tag scores corresponding to the K financial tags respectively; and replacing the financial label score corresponding to the financial label smaller than the preset score threshold value with a preset value to obtain new financial label scores corresponding to the K financial labels respectively.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: removing financial tags with probability smaller than a preset probability threshold from the K financial tags; and sorting other financial tags except for the financial tag with the probability smaller than the preset probability threshold value in the K financial tags to obtain the sorting result.
Optionally, the above computer program product is further adapted to execute a program initialized with the method steps of: receiving a grading value of the financial product included in the financial product recommendation result aiming at the target account; and removing financial products with score values smaller than a preset score threshold value from the financial products included in the financial product recommendation result to obtain a new financial product recommendation result.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the present application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. which are within the spirit and principles of the present application are intended to be included within the scope of the claims of the present application.

Claims (11)

1. A financial product recommendation method, comprising:
acquiring first account feature data of a target account;
based on the first account characteristic data, a random forest financial management prediction model is adopted to obtain financial management labels respectively output by N random trees and label scores corresponding to the financial management labels respectively output by the N random trees, wherein the random forest financial management prediction model is obtained by training the financial management labels respectively corresponding to the historical financial management behaviors of M accounts based on account characteristic data of the M accounts, M, N is an integer greater than or equal to 2, and the random forest financial management prediction model comprises the N random trees;
obtaining financial label scores corresponding to K financial labels corresponding to the target account according to the financial labels respectively output by the N random trees and label scores corresponding to the financial labels respectively output by the N random trees, wherein K is an integer greater than or equal to 2;
and determining a financial product recommendation result of the target account based on the financial label scores respectively corresponding to the K financial labels.
2. The method of claim 1, wherein before the obtaining, based on the first account feature data of the target account, the financial tags respectively output by the N random trees and the tag scores corresponding to the financial tags respectively output by the N random trees by using a random forest financial prediction model, the method further comprises:
Acquiring account characteristic data of the M accounts and financial tags corresponding to historical financial behaviors of the M accounts respectively;
based on account characteristic data of the M accounts and financial labels corresponding to historical financial behaviors of the M accounts, performing L rounds of node division according to the financial labels to generate the N random trees, wherein L is an integer greater than or equal to 2;
and obtaining the random forest financial prediction model according to the N random trees.
3. The method of claim 2, wherein generating the N random trees based on the account feature data of the M accounts and financial tags corresponding to the historical financial behaviors of the M accounts, respectively, includes:
for any current round of node division in the L rounds of node division, performing the current round of node division according to the financial tags respectively corresponding to the account feature data of the M accounts and the historical financial behaviors of the M accounts until the number of the financial tags and the number of the account feature data included in each leaf node obtained by the current round of node division are in a corresponding preset number range:
Dividing account characteristic data corresponding to the dividing nodes of the previous round of the current round into two leaf nodes according to financial labels respectively corresponding to the historical financial behaviors of the M accounts, under the condition that the dividing node of the previous round is a root node, account feature data corresponding to the dividing node of the previous round is account feature data of the M accounts;
determining the types of financial tags respectively included in the two leaf nodes and the number of each financial tag;
sorting the financial tags respectively included in the two leaf nodes according to the number of each financial tag to obtain a first sorting result;
and re-dividing the account feature data corresponding to the two leaf nodes according to the first sorting result to obtain new two leaf nodes and a second sorting result until the first sorting result is the same as the second sorting result.
4. A method according to claim 3, characterized in that the method further comprises:
and under the condition that the second sorting result and the first sorting result are different, taking the new two leaf nodes as the two leaf nodes and the second sorting result as the first sorting result, repeatedly executing the node division on account feature data respectively included by the two leaf nodes according to the first sorting result to obtain the new two leaf nodes, and operating the second sorting result until the second sorting result and the first sorting result are the same.
5. The method of claim 1, wherein the obtaining, based on the first account feature data of the target account, the financial tags respectively output by the N random trees and tag scores corresponding to the financial tags respectively output by the N random trees by using a random forest financial prediction model includes:
based on the first account characteristic data of the target account, adopting the random forest financial management prediction model to obtain financial tags output by at least one leaf node corresponding to the N random trees respectively and tag scores corresponding to the financial tags output by the at least one leaf node;
and obtaining financial labels respectively output by the N random trees and label scores corresponding to the financial labels respectively output by the N random trees according to the financial labels respectively output by at least one leaf node and the first label scores corresponding to the financial labels output by the at least one leaf node.
6. The method of claim 1, wherein the determining the financial product recommendation of the target account based on the financial tag scores corresponding to the K financial tags, respectively, comprises:
Updating the financial tag scores corresponding to the K financial tags respectively based on a preset score threshold to obtain new financial tag scores corresponding to the K financial tags respectively;
normalizing the new financial label scores corresponding to the K financial labels respectively to obtain the financial label probabilities corresponding to the K financial labels respectively;
sorting the K financial tags according to the probability of the financial tags corresponding to the K financial tags respectively to obtain a third sorting result;
and determining the financial product recommendation result of the target account according to the third sorting result.
7. The method of claim 6, wherein updating the financial tag scores corresponding to the K financial tags respectively based on a preset score threshold value, to obtain new financial tag scores corresponding to the K financial tags corresponding to the target account respectively comprises:
determining financial tags smaller than the preset score threshold in the financial tag scores corresponding to the K financial tags respectively;
and replacing the financial label scores corresponding to the financial labels smaller than the preset score threshold value with preset values to obtain new financial label scores corresponding to the K financial labels respectively.
8. The method of claim 7, wherein the sorting the K financial tags according to the probability of the financial tag corresponding to the K financial tags, to obtain a sorting result, includes:
removing financial tags with probability smaller than a preset probability threshold from the K financial tags;
and sorting other financial tags except for the financial tag with the probability smaller than the preset probability threshold value in the K financial tags to obtain the sorting result.
9. The method according to any one of claims 1 to 8, wherein after the determining the financial product recommendation of the target account based on the financial tag scores corresponding to the K financial tags, respectively, the method further comprises:
receiving a grading value of the financial product included in the financial product recommendation result aiming at the target account;
and removing financial products with score values smaller than a preset score threshold value from the financial products included in the financial product recommendation result to obtain a new financial product recommendation result.
10. A financial product recommendation device, comprising:
The first acquisition module is used for acquiring first account characteristic data of the target account;
the prediction module is used for obtaining financial tags respectively output by N random trees and tag scores corresponding to the financial tags respectively output by the N random trees based on the first account characteristic data, wherein the random forest financial prediction model is obtained by training the financial tags respectively corresponding to the historical financial behaviors of M accounts based on the account characteristic data of the M accounts, M, N is an integer greater than or equal to 2, and the random forest financial prediction model comprises the N random trees;
the second acquisition module is used for acquiring financial tag scores corresponding to K financial tags corresponding to the target account according to the financial tags respectively output by the N random trees and the tag scores corresponding to the financial tags respectively output by the N random trees, wherein K is an integer greater than or equal to 2;
and the determining module is used for determining the financial product recommendation result of the target account based on the financial label scores respectively corresponding to the K financial labels.
11. A non-volatile storage medium storing a plurality of instructions adapted to be loaded by a processor and to perform the financial product recommendation method of any one of claims 1 to 9.
CN202310539421.0A 2023-05-12 2023-05-12 Financial product recommendation method, device and storage medium Pending CN116362895A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310539421.0A CN116362895A (en) 2023-05-12 2023-05-12 Financial product recommendation method, device and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310539421.0A CN116362895A (en) 2023-05-12 2023-05-12 Financial product recommendation method, device and storage medium

Publications (1)

Publication Number Publication Date
CN116362895A true CN116362895A (en) 2023-06-30

Family

ID=86937626

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310539421.0A Pending CN116362895A (en) 2023-05-12 2023-05-12 Financial product recommendation method, device and storage medium

Country Status (1)

Country Link
CN (1) CN116362895A (en)

Similar Documents

Publication Publication Date Title
CN105718490A (en) Method and device for updating classifying model
CN106875206A (en) Acquisition of information, assessment, questionnaire method, device and server
US8984022B1 (en) Automating growth and evaluation of segmentation trees
CN110930038A (en) Loan demand identification method, loan demand identification device, loan demand identification terminal and loan demand identification storage medium
CN115170294A (en) Client classification method and device and server
CN116450951A (en) Service recommendation method and device, storage medium and electronic equipment
CN114896506A (en) Product recommendation method, device, equipment and storage medium
CN113450158A (en) Bank activity information pushing method and device
US20210004722A1 (en) Prediction task assistance apparatus and prediction task assistance method
US11120503B2 (en) Analysis and visual presentation of dataset components
CN116800831A (en) Service data pushing method, device, storage medium and processor
CN116630020A (en) Risk assessment method and device, storage medium and electronic equipment
CN116362895A (en) Financial product recommendation method, device and storage medium
US20220067460A1 (en) Variance Characterization Based on Feature Contribution
CN115186101A (en) Investment management back-end system, method, equipment and storage medium
CN114493279A (en) Workflow task prediction method and device, storage medium and electronic equipment
CN113240513A (en) Method for determining user credit line and related device
CN113870007A (en) Product recommendation method, device, equipment and medium
CN111461863A (en) Data processing method and device, computer equipment and storage medium
CN111932018B (en) Bank business performance contribution information prediction method and device
US11803915B1 (en) Computing system for adaptive investment recommendations
CN116051169A (en) Evaluation method, evaluation device, nonvolatile storage medium and electronic equipment
RU2710830C1 (en) Method for automated provision of recommendations on making investment-relevant decisions on the market of digital assets and device for its implementation
Barrueta-Meza et al. Predictive Model to Determine Customer Desertion in Peruvian Banking Entities
CN113205338A (en) Foreign exchange service processing method and device based on artificial intelligence

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