CN117592859A

CN117592859A - LSTM-based customer financial risk assessment method, device and equipment

Info

Publication number: CN117592859A
Application number: CN202311648909.3A
Authority: CN
Inventors: 崔亚昆; 杨超
Original assignee: Agricultural Bank of China
Current assignee: Agricultural Bank of China
Priority date: 2023-12-04
Filing date: 2023-12-04
Publication date: 2024-02-23

Abstract

The application provides a customer financial risk assessment method, device and equipment based on LSTM, which relate to the field of financial risk assessment, and are used for acquiring assessment data corresponding to influence factors of each assessment type in multiple assessment types of customers, wherein the assessment data comprises test question answering results, answering consumption time and request prompting times, and inputting the assessment data into a risk assessment model corresponding to each assessment type in multiple assessment types to obtain risk assessment results of the customers in each assessment type; according to the risk assessment results of various assessment types and the basic information of the clients, determining the risk assessment image of the clients, accurately determining the risk assessment results of the clients aiming at the target financial products based on the risk assessment image and the risk grade image of the target financial products, reducing the deviation between the risk assessment results of the clients and the risk bearing capacity of the clients purchasing the financial products, and improving the matching degree of the purchased financial products and the risk bearing capacity of the clients.

Description

LSTM-based customer financial risk assessment method, device and equipment

Technical Field

The application relates to the field of financial risk assessment, in particular to a customer financial risk assessment method, device and equipment based on LSTM.

Background

The risk assessment result reflects the risk bearing capacity of the customer when the customer purchases the financial product, the online banking is one of the main channels for selling the financial product to the customer, and the risk assessment system accurately marks the risk bearing capacity of the customer, thereby being beneficial to finding potential customers and reducing economic losses caused by insufficient bearing capacity for the customers.

In the related art, the result of the evaluation of the financial risk of the customer is usually determined by counting the accuracy of answering the questions by the customer, wherein the questions generally include information such as the age, income, region and intention of the customer to purchase the financial product.

However, the risk assessment result of the customer financial management obtained by adopting the above technical scheme is generally greatly deviated from the risk bearing capacity of the customer for purchasing the financial management product, so that the purchased financial management product is not matched with the risk bearing capacity of the customer.

Disclosure of Invention

The application provides a customer financial risk assessment method, device and equipment based on LSTM (least squares), which are used for solving the problem that the risk assessment result of customer financial risk obtained by the related technical scheme is generally greatly deviated from the risk bearing capacity of purchasing financial products by customers, so that the purchased financial products are not matched with the risk bearing capacity of the customers.

In a first aspect, the present application provides an LSTM-based client financial risk assessment method, including:

acquiring evaluation data of a customer aiming at influence factors of each evaluation type in a plurality of evaluation types, wherein the plurality of evaluation types comprise financial knowledge storage conditions, inauguration investment experience, future dominant income expectations and investment attitudes, and the evaluation data comprise test question answering results, answering consumption time and request prompting times;

inputting the evaluation data of the influence factors of the evaluation types into a risk evaluation model corresponding to the evaluation types aiming at the evaluation data of the influence factors of each evaluation type in a plurality of evaluation types to obtain a risk evaluation result of a client aiming at the evaluation types, wherein the risk evaluation model is a model based on a long-short-term memory network LSTM;

determining a risk evaluation figure of the client according to a risk evaluation result of the client aiming at each evaluation type in a plurality of evaluation types and basic information of the client, wherein the basic information comprises occupation, annual income and asset conditions;

and determining financial risk assessment results of the clients aiming at the target financial products based on the risk assessment images and the risk level images of the target financial products.

In one possible implementation manner, the risk assessment model includes an encoder of a self-encoder and an LSTM model, and the step of inputting the assessment data of the influence factors of the assessment type into the risk assessment model of the corresponding assessment type to obtain a risk assessment result of the customer for the assessment type includes: performing single-heat coding on the evaluation data of the influence factors of the evaluation type to obtain first coded data; inputting the first encoded data into an encoder for dimension reduction processing to obtain second encoded data, wherein the dimension of the second encoded data is smaller than that of the first encoded data; and inputting the second coded data into the LSTM model for risk assessment, and obtaining a risk assessment result of a customer aiming at an assessment type.

In one possible implementation, determining the financial risk assessment result of the client for the target financial product based on the risk assessment image and the risk level image of the target financial product includes: determining whether the customer meets a risk level representation of the target financial product based on the risk assessment representation; if the client meets the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is passed, and outputting a purchase page of the target financial product or a target financial product list containing the target financial product; if the client does not meet the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is failed, and returning to execute the step of inputting the assessment data of the influence factors of the assessment type into the risk assessment model of the corresponding assessment type to obtain the risk assessment result of the client for the assessment type.

In one possible implementation, the risk assessment image includes a capability of the customer for each factor corresponding to the assessment type, and determining whether the customer meets the risk level image of the target financial product based on the risk assessment image includes: determining whether the bearing capacity of the client for each factor corresponding to the evaluation type accords with the risk level portrait of the target financial product; if the bearing capacity of the client for each factor corresponding to the evaluation type accords with the risk level portrait of the target financial product, determining that the client meets the risk level portrait of the target financial product; if the bearing capacity of the client corresponding to any factor of the evaluation type does not meet the risk level portrait of the target financial product, determining that the client does not meet the risk level portrait of the target financial product.

In one possible implementation manner, the evaluation data of the influence factors of the evaluation type are obtained by digitally signing the acquired evaluation data of the influence factors of the evaluation type at the client, and the evaluation data of the influence factors of the evaluation type are input into a risk evaluation model corresponding to the evaluation type to obtain a risk evaluation result of the client for the evaluation type, and the method includes: checking labels are carried out on the evaluation data of the influence factors of the evaluation types; after the verification passes, the evaluation data of the influence factors of the evaluation type are input into a risk evaluation model corresponding to the evaluation type, and a risk evaluation result of the client aiming at the evaluation type is obtained.

In one possible embodiment, the evaluation data is obtained by: after the customer selects the target financial product through man-machine interaction, if the customer is determined to not meet the risk grade portrait of the target financial product according to the current risk evaluation result of the customer, the evaluation test questions of each evaluation type in a plurality of evaluation types are displayed so as to acquire evaluation data of influence factors corresponding to each evaluation type.

In one possible implementation, the evaluation data of the influence factors of the evaluation type are also recorded in the database, and are used for updating the risk evaluation model of the corresponding evaluation type after invalid data are removed.

In a second aspect, the present application provides an LSTM-based customer financial risk assessment device, including:

the acquisition module is used for acquiring evaluation data of influence factors of clients aiming at each evaluation type in a plurality of evaluation types, wherein the plurality of evaluation types comprise financial knowledge storage conditions, inauguration investment experience, future dominant income expectations and investment attitudes, and the evaluation data comprise test question answering results, answering consumption time and request prompting times;

the evaluation module is used for inputting the evaluation data of the influence factors of the evaluation types into a risk evaluation model corresponding to the evaluation types aiming at the evaluation data of the influence factors of each evaluation type in the plurality of evaluation types to obtain a risk evaluation result of a client aiming at the evaluation types, wherein the risk evaluation model is a model based on a long-short-term memory network LSTM;

the first determining module is used for determining a risk assessment portrait of the client according to a risk assessment result of the client for each of a plurality of assessment types and basic information of the client, wherein the basic information comprises occupation, annual income and asset conditions;

and the second determining module is used for determining financial risk assessment results of clients aiming at the target financial products based on the risk assessment figures and the risk level figures of the target financial products.

In a third aspect, the present application provides an electronic device, comprising: a processor, a memory communicatively coupled to the processor;

a memory for storing computer-executable instructions;

a processor for executing computer-executable instructions stored in a memory to implement the method of any one of the first aspects.

In a fourth aspect, the present application provides a computer-readable storage medium having stored therein computer-executable instructions for performing the method of any of the first aspects when the computer-executable instructions are executed.

In a fifth aspect, the present application provides a computer program product comprising a computer program which, when executed, implements the method of any one of the first aspects.

According to the LSTM-based customer financial risk assessment method, device and equipment, the assessment data corresponding to the influence factors of each assessment type in the plurality of assessment types are obtained, the assessment data comprise test question answer results, answer consumption time and request prompt times, the assessment data are input into the LSTM-based risk assessment model corresponding to the assessment types, the influence of the plurality of influence factors in the risk bearing capacity, the mutual influence among different influence factors and the answer results, answer consumption time and request prompt times corresponding to the influence factors are fully considered, the risk assessment results of the customers for the influence factors of the plurality of assessment types are obtained, the risk assessment results of the customers are further determined according to the risk assessment results of the plurality of assessment types and the customer basic information, and the risk assessment results of the customers for the target financial products are accurately determined based on the risk assessment images and the risk grades of the target financial products, so that the deviation of the risk assessment results of the customers and the risk bearing capacity of the customer purchasing the financial products is reduced, and the risk bearing capacity of the customer purchasing the financial products is further improved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.

FIG. 1 is a schematic diagram of the output of a single thermal encoding input LSTM model provided in an exemplary embodiment of the present application;

FIG. 2 is a schematic diagram of a self-encoder structure provided in an exemplary embodiment of the present application;

fig. 3 is an application scenario schematic diagram of an LSTM-based client financial risk assessment method according to an exemplary embodiment of the present application;

FIG. 4 is a flow chart of a customer financial risk assessment method based on LSTM according to an exemplary embodiment of the present application;

fig. 5 is a schematic structural diagram of an LSTM unit provided in an exemplary embodiment of the present application;

FIG. 6 is a schematic diagram of model training of the financial knowledge reserve condition assessment type provided by an exemplary embodiment of the present application;

FIG. 7 is a schematic diagram of an LSTM-based customer financial risk assessment system according to an exemplary embodiment of the present application;

FIG. 8 is another flow chart of the LSTM-based customer financial risk assessment method according to the exemplary embodiment of the present application;

FIG. 9 is a schematic diagram of a structure of an LSTM-based customer financial risk assessment device according to an exemplary embodiment of the present application;

Fig. 10 is a schematic structural diagram of an electronic device according to an exemplary embodiment of the present application.

Specific embodiments thereof have been shown by way of example in the drawings and will herein be described in more detail. These drawings and the written description are not intended to limit the scope of the inventive concepts in any way, but to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application as detailed in the accompanying claims.

The terms first, second and the like in the description and in the claims, 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 such that embodiments of the present application described herein may be capable of operation in sequences other than those illustrated or described herein, for example. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, 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, article, or apparatus.

First, some terms related to the present application will be explained:

long-term memory network: a Long Short-Term Memory (LSTM) is used as a kind of cyclic neural network (Recurrent Neural Network, RNN) and its hidden layer state is determined by the input of the current time and the activation state of the previous time.

Single heat coding: one-Hot Encoding (One-Hot Encoding), also called One-bit effective Encoding, wherein each sequence of the One-Hot Encoding has only One 1, the other bits are 0, the former T bit marks the condition of question answering, the latter T is the condition of wrong answer of the marked question, T is a positive integer greater than 1, for example, one record is that the question answering corresponding to the third knowledge point is correct, then the T+3rd element in the input sequence is 1, and the rest is 0; if the question corresponding to the third knowledge point is wrong, the third bit in the input sequence is 1, and the rest is 0, so that each answer record can be represented by a sequence with the same length. Illustratively, fig. 1 is a schematic diagram of output of inputting a single thermal code into an LSTM model provided in an exemplary embodiment of the present application. As shown in fig. 1, x ₁ 、x ₂ 、x ₃ 、x _T Representation of input single-hot code, h ₀ 、h ₁ 、h ₂ 、h ₃ 、h _T For LSTM model, when t=3, y ₁ 、y ₂ 、y ₃ 、y _T The outputs corresponding to the one-hot codes, the outputs corresponding to the scores, e.g., 0.6, 0.5, and 0.8 in fig. 1; err is the input sequence of the single-hot code corresponding to the question answering error corresponding to the third knowledge point, and Cor is the input sequence of the single-hot code corresponding to the question answering correct corresponding to the third knowledge point.

The self-encoder is an unsupervised neural network model, and uses the nonlinear fitting characteristic of the neural network to mine the implicit relation between the input data characteristics so as to realize the characteristic dimension reduction. The self-encoder consists of an encoder and a decoder, wherein the encoder is responsible for compressing input characteristics to a certain length, and the decoder is responsible for restoring the compression result to a state when data is input. In the training process, the loss function is the difference between the original input data and the decoding result, and model parameters are continuously updated through reverse transfer, so that the accuracy of the self-encoder is gradually improved. Illustratively, fig. 2 is a schematic diagram of a self-encoder structure provided in an exemplary embodiment of the present application. As shown in FIG. 2, where N is the number of secondary influencing factors (e.g. the number of points of financial knowledge or pieces of investment experience), M is the discrete range of values when responding, P is the range of values when requesting the number of prompts, input_layer is the input layer, input_size is the length of the input vector of the input layer, hidden_layer is the hidden layer, N1 is the length of the input vector of the hidden layer, output_layer is the output layer, W represents the set of conversion parameters from the input layer to the hidden layer, W ^T Representing a set of hidden layer to output layer conversion parameters, W ^T Corresponding to the transpose of W, x is the input vector of the input layer, f (x) is the encoder function of the self-encoder, g (f (x)) is the decoder function of the self-encoder, and loss is the loss function of the input vector x and decoded output vector. The encoder is connected with the decoder when the model is trained, and the encoder is only connected with the risk assessment model when the model is in online service.

Learning process data: the data which can reflect the answering situation in the answering and evaluating process not only comprises the answering result, but also comprises the answering time consumption, the number of request prompts and the like, because the answering result cannot reflect the strength of a client on a certain item of capability more accurately, and the risk bearing capability of the client cannot be reflected accurately.

In the related technology, the correct rate of the answer test questions of the clients is counted to determine the financial risk assessment result of the clients, the number of the related test questions is small, the investigation factors are not comprehensive enough, only the basic information such as the ages, incomes and areas of the clients and the intention of the clients to purchase financial products are considered, the knowledge of the clients to manage the financial is not investigated, such as basic knowledge such as yield calculation and financial strategies, and the content of the assessment test questions is not progressive and has strong counterfeitability, so the real situation cannot be reflected; and the requirements of different financial products on various risk bearing capacity evaluation indexes of the clients are different, if only one statistical score is used instead of a vector consisting of different indexes, the matching degree between the financial products and the clients can be influenced. In addition, the answering process only considers whether the answering result is correct or not, and the answering time and the request prompting times are not examined, so that the obtained client financial risk assessment result is generally in larger deviation with the risk bearing capacity of the client for purchasing the financial product, and the purchased financial product is not matched with the risk bearing capacity of the client.

Since the risk tolerance is determined by four primary factors (i.e., financial knowledge reserve, risk investment experience, future dominant income expectations, and investment attitudes) together, and each primary factor is determined by scores of multiple secondary factors, a more accurate total risk tolerance score can be obtained only by reasonably quantifying the scores of the secondary factors during evaluation. In order to solve the above problems, the embodiment of the application provides a customer financial risk assessment scheme based on LSTM, by subdividing the assessment content into five modules of customer basic information, financial knowledge storage condition, risk investment experience, future available income expectation and investment attitude, wherein the latter four modules reflect the risk bearing capacity of customers purchasing financial products, each module is represented by a plurality of sub-elements together, in the assessment, each assessment question can examine one or more secondary factors, and by refining the investigation index, the risk bearing capacity of customers can be reflected more realistically; in addition, the answer result cannot uniquely reflect the real situation of the customer, the answer is correct, the time-consuming time is certainly lower than the time-consuming time on the capability, for example, when the situation of evaluating financial knowledge storage is tested, the answer is correct, the financial knowledge of the customer which takes 20 seconds is certainly more complete than the financial knowledge of the customer which takes 1 minute, so that the answer result of each module is inspected during the test, the corresponding answer consumption time and the request prompting times are inspected, the LSTM model is introduced to complete the test of the risk capability, the interaction among different subelements is quantized through the LSTM model, the obtained risk test result of the customer aiming at the test type is more accurate, the deviation of the risk test result of the customer financial risk and the risk bearing capability of the customer for purchasing the financial product is reduced, and the matching degree of the purchased financial product and the risk bearing capability of the customer is further improved.

Fig. 3 is an application scenario schematic diagram of the LSTM-based client financial risk assessment method according to an exemplary embodiment of the present application. As shown in fig. 3, the application scenario includes a client 31 and a server 32, where the number of clients 31 may be at least one. In practical application, when detecting the evaluation data of the influence factors of each evaluation type in the multiple evaluation types submitted by the client through the client 31, the server 32 executes the client financial risk evaluation method based on the LSTM provided by the application to obtain the financial risk evaluation result of the client for the target financial product.

It should be noted that the server 32 may be replaced by a server cluster or other computing device with a certain computing power. The client 31 may be a computer, a mobile phone, a notebook or a personal digital assistant (Personal Digital Assistant, PDA for short), etc.

The customer financial risk assessment method based on LSTM according to an exemplary embodiment of the present application is described below with reference to fig. 4 in conjunction with the application scenario of fig. 3. It should be noted that the above application scenario is only shown for the convenience of understanding the spirit and principles of the present application, and the embodiments of the present application are not limited by the application scenario shown in fig. 3.

Fig. 4 is a flow chart of a customer financial risk assessment method based on LSTM according to an exemplary embodiment of the present application. As shown in fig. 4, the client financial risk assessment method based on LSTM in the embodiment of the present application includes the following steps:

s401, acquiring evaluation data of influence factors of clients aiming at each of a plurality of evaluation types, wherein the plurality of evaluation types comprise financial knowledge storage conditions, inauguration investment experience, future dominant income expectations and investment attitudes, and the evaluation data comprise test question answering results, answering consumption time and request prompting times.

In this step, as shown in fig. 3, the customer sequentially performs the solution of the financial knowledge reserve situation, the inauguration investment experience, the future dominant income expectation and the investment attitude evaluation questions through the client 31, each of which represents the investigation of one secondary factor (for example, the investment product yield calculation method, the importance degree of the inauguration investment product in the asset configuration, etc.), and the evaluation questions between the respective modules are not mixed. Illustratively, table 1 is a test question data model provided by the exemplary embodiment of the present application, and table 2 is a two-level factor data model provided by the exemplary embodiment of the present application.

TABLE 1

Field name	Field type	Main key	External key	Whether or not it can be empty	Field description
						ques_id	int	√	×	NOT NULL	Test question numbering
ques_content	varchar	×	×	NOT NULL	Content of test questions
						ques_sec_id	varchar	×	×	NOT NULL	Number of secondary factors examined by test questions

TABLE 2

Field name	Field type	Main key	External key	Whether or not it can be empty	Field description
						factor_sec_id	int	√	×	NOT NULL	Secondary influencing factor numbering
factor_fir_id	varchar	×	×	NOT NULL	Corresponding to the number of the first-level influencing factors
						factor_sec_name	varchar	×	×	NOT NULL	Second order influence factor name

As shown in tables 1 and 2, each test question examines one or more secondary factors, each corresponding to a unique primary factor.

In the evaluation process, the server 32 starts a timer, records the answer time of each test question in each evaluation type, and starts a counter to record the request prompting times of each test question in each evaluation type; after the evaluation is completed, the client submits the answer result of the test questions through the client 31. Correspondingly, the server 32 obtains evaluation data of influence factors of each evaluation type in the customer for the financial knowledge reserve situation, the inauguration investment experience, the future dominant income expectation and the investment attitude, and the evaluation data includes test question answering results, answering consuming time and request prompting times.

S402, inputting the evaluation data of the influence factors of the evaluation types into a risk evaluation model of the corresponding evaluation type aiming at the evaluation data of the influence factors of each evaluation type in a plurality of evaluation types to obtain a risk evaluation result of a client aiming at the evaluation type, wherein the risk evaluation model is based on an LSTM (least squares) model.

The method comprises the steps of inputting evaluation data of financial knowledge reserve conditions into a first service model based on LSTM to obtain risk evaluation results of customers aiming at the financial knowledge reserve conditions, wherein the first service model based on LSTM is a pre-trained risk evaluation model for carrying out the financial knowledge reserve conditions; inputting the evaluation data of the inauguration investment experience into a second service model based on the LSTM to obtain a inauguration evaluation result of the client aiming at the inauguration investment experience, wherein the second service model based on the LSTM is a inauguration evaluation model which is trained in advance and used for carrying out the inauguration investment experience; inputting the future dominant revenue expected evaluation data into a third service model based on LSTM to obtain risk evaluation results of customers aiming at the future dominant revenue expected, wherein the third service model based on LSTM is a pre-trained risk evaluation model for future dominant revenue expected; inputting the evaluation data of the investment attitude into a fourth service model based on the LSTM to obtain a risk evaluation result of the client aiming at the investment attitude, wherein the fourth service model based on the LSTM is a pre-trained risk evaluation model for carrying out the investment attitude.

Correspondingly, after each type of evaluation is completed, the corresponding evaluation result is not simply superposition of several influencing factors, but is a 1×n one-dimensional vector, N is the number of each secondary influencing factor, each element represents the score of the client on the corresponding secondary factor, and the higher the score is, the stronger the bearing capacity of the client on the secondary factor is represented.

S403, determining risk assessment figures of the clients according to risk assessment results of the clients aiming at each of a plurality of assessment types and basic information of the clients, wherein the basic information comprises occupation, annual income and asset conditions.

In this step, based on the evaluation results of the financial knowledge reserve situation, the inauguration investment experience, the future dominant income expectation and the investment attitude, namely, 4 one-dimensional vectors, the weighted average of the one-dimensional vectors obtained by evaluation of the financial knowledge reserve situation is taken as the score of the financial knowledge reserve situation, the weighted average of the one-dimensional vectors obtained by evaluation of the inauguration investment experience is taken as the score of the inauguration investment experience, the weighted average of the one-dimensional vectors obtained by evaluation of the future dominant income expectation is taken as the score of the future dominant income expectation, and the weighted average of the one-dimensional vectors obtained by evaluation of the investment attitude is taken as the score of the investment attitude. Further, a risk assessment representation of the customer is determined based on the customer's basic information, scores for financial knowledge reserves, scores for risk investment experience, scores for future dominant revenue expectations, and scores for investment attitudes.

Considering that the risk assessment is completed only through a statistical mode in the related technology, the obtained wind assessment result is single, and the recommending function of the financial product cannot be supported, after each time of the assessment is completed, the method and the device obtain 4 one-dimensional vectors, namely 4 XN state matrixes, can intuitively reflect the risk bearing capacity of a customer, can be used for recommending matched financial products for the customer, and further provide a certain guarantee for the recommending function expansion of the financial module.

S404, determining financial risk assessment results of clients aiming at target financial products based on the risk assessment figures and the risk level figures of the target financial products.

The risk assessment image of the client and the risk level image of the target financial product are compared according to a preset rule, and if the risk assessment image of the client meets the requirement of the risk level image of the target financial product, the financial risk assessment result of the client for the target financial product is determined to be passed; if the risk assessment image of the client does not meet the requirements of the risk level image of the target financial product, determining that the financial risk assessment result of the client aiming at the target financial product is not passed.

According to the customer financial risk assessment method based on the LSTM, the customer is provided with the evaluation data aiming at the influence factors of each evaluation type in the plurality of evaluation types, the evaluation data are input into the risk evaluation model based on the LSTM, the influence factors of the plurality of types in the risk bearing capacity, the mutual influence among the influence factors and the response result, the response consumption time and the request prompting times corresponding to the influence factors are fully considered, the obtained customer is more accurate in the risk evaluation result aiming at the evaluation types, deviation of the risk bearing capacity of the customer financial risk evaluation result and the customer purchasing financial products is reduced, and matching degree of the purchased financial products and the risk bearing capacity of the customer is further improved.

In some embodiments, the assessment data is obtained by: after the customer selects the target financial product through man-machine interaction, if the customer is determined to not meet the risk grade portrait of the target financial product according to the current risk evaluation result of the customer, the evaluation test questions of each evaluation type in a plurality of evaluation types are displayed so as to acquire evaluation data of influence factors corresponding to each evaluation type.

As shown in fig. 3, for example, a client accesses an internet banking service through a client 31, enters a financial management module to select a target financial product and submits a purchase request, and correspondingly, a server 32 responds to the purchase request to detect whether the client has performed financial risk assessment or whether the current risk assessment result of the client is matched with a risk level portrait of the target financial product, and if the client does not perform financial risk assessment or the current risk assessment result of the client is not matched with the risk level portrait of the target financial product, sequentially displays assessment questions of each assessment type in the client 31, and collects answer results, answer consumption time and request prompt times of the questions so as to collect assessment data of influence factors corresponding to each assessment type.

In some embodiments, determining the financial risk assessment result for the target financial product by the customer based on the risk assessment image and the risk level image of the target financial product comprises: determining whether the customer meets a risk level representation of the target financial product based on the risk assessment representation; if the client meets the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is passed, and outputting a purchase page of the target financial product or a target financial product list containing the target financial product; if the client does not meet the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is failed, and returning to execute the step of inputting the assessment data of the influence factors of the assessment type into the risk assessment model of the corresponding assessment type to obtain the risk assessment result of the client for the assessment type.

For example, as shown in fig. 3, if the target financial product of the purchase request is selected and submitted by the client 31 as the a product, the server 32 responds to the purchase request for the a product, and determines whether the client satisfies the risk level portrait of the a product based on the risk assessment portrait of the client carried in the purchase request or based on the risk assessment portrait of the client stored in the server 32; if the customer meets the risk level portrait of the A product, determining that the financial risk assessment result of the customer for the A product is passed, and outputting a purchase page of the A product or a target financial product list containing the A product; if the customer does not meet the risk level portrait of the A product, determining that the financial risk assessment result of the customer for the A product is failed, sequentially displaying assessment questions of each assessment type in the client 31, and collecting answer results, answer consumption time and request prompting times of each question again to collect assessment data of influence factors corresponding to each assessment type; further, the server 32 returns to execute the step of inputting the evaluation data of the influence factors of the evaluation type into the risk evaluation model of the corresponding evaluation type, and obtaining the risk evaluation result of the client for the evaluation type. If the customer does not satisfy the risk level portrait of the product a, a financial product list matching the risk assessment portrait of the customer may be displayed on the client 31, and the customer may select to purchase other financial products meeting personal expectations based on the financial product list.

Based on the above embodiments, in some embodiments, the risk assessment image includes the bearing capability of the customer for each factor corresponding to the assessment type, and determining, based on the risk assessment image, whether the customer meets the risk level image of the target financial product includes: determining whether the bearing capacity of the client for each factor corresponding to the evaluation type accords with the risk level portrait of the target financial product; if the bearing capacity of the client for each factor corresponding to the evaluation type accords with the risk level portrait of the target financial product, determining that the client meets the risk level portrait of the target financial product; if the bearing capacity of the client corresponding to any factor of the evaluation type does not meet the risk level portrait of the target financial product, determining that the client does not meet the risk level portrait of the target financial product.

Illustratively, the risk level representation of the target financial product includes a score threshold for financial knowledge reserve, a score threshold for inauguration investment experience, a score threshold for future dominant revenue expectations, and a score threshold for investment attitude. Correspondingly, if the score of the financial knowledge reserve condition, the score of the risk investment experience, the score of the future dominant income expectation and the score of the investment attitude in the risk assessment image of the client are all greater than or equal to the corresponding threshold values, namely, the bearing capacity of the client for each factor corresponding to the assessment type accords with the risk grade image of the target financial product, determining that the financial risk assessment result of the client for the target financial product is passed; if any one of the score of the financial knowledge reserve condition, the score of the risk investment experience, the score of the future dominant income expectation and the score of the investment attitude in the risk assessment image of the client is smaller than the corresponding threshold, namely, the bearing capacity of the client for any factor corresponding to the assessment type does not accord with the risk grade image of the target financial product, determining that the financial risk assessment result of the client for the target financial product is not passed.

In some embodiments, the risk assessment model includes an encoder of a self-encoder and an LSTM model, and inputting assessment data of an impact factor of an assessment type into the risk assessment model of a corresponding assessment type to obtain a risk assessment result of a customer for the assessment type, including: performing single-heat coding on the evaluation data of the influence factors of the evaluation type to obtain first coded data; inputting the first encoded data into an encoder for dimension reduction processing to obtain second encoded data, wherein the dimension of the second encoded data is smaller than that of the first encoded data; and inputting the second coded data into the LSTM model for risk assessment, and obtaining a risk assessment result of a customer aiming at an assessment type.

Correspondingly, as the LSTM model has a memory function on a time sequence, the memory capability of a person can be simulated, and the model is closer to a real answer scene of a customer, for example, when evaluating the situation of financial knowledge storage, if a plurality of questions are continuously answered, the score is higher than that when only one question is answered. Fig. 5 is a schematic structural diagram of an LSTM unit according to an exemplary embodiment of the present application, where an LSTM model is composed of a plurality of LSTM units. As shown in fig. 5, X (t) represents an input of the LSTM cell at time t, h (t) represents an output of the LSTM cell at time t, C (t) represents another output of the LSTM cell at time t, and h (t-1) represents an output of the LSTM cell at time t-1, which is also an input of the LSTM cell at time t.

Accordingly, after the evaluation data of the influence factors of the clients aiming at each of the multiple evaluation types is obtained, because the evaluation data comprises the answer result of the test questions, the answer consumption time and the request prompting times, the number of the sub factors to be considered is large, if the evaluation data is directly subjected to the single-heat coding treatment and then is input into the LSTM model, the performance of the model can be influenced. Therefore, in order to minimize loss of important information, an encoder introduced from the encoder performs dimension reduction on the input to improve the accuracy of the risk assessment model.

For example, if the evaluation of the financial knowledge reserve condition is currently performed, if the number of the evaluation of the financial knowledge points is N, N is the number of the secondary influencing factors, the single-heat encoding length of the answer result is 2*N; because the response time is continuous, discretization is needed to be carried out, a fixed time interval is defined, if the response time is within 0-Mmin, the time interval can be set to be 1min, and the corresponding single-heat coding length of the response time is 2*M; if the request prompting times range is 0-P times, the single thermal coding length of the request prompting times is 2*P, and the corresponding model input length is 2×N+2×M+2×P. For example, the knowledge point of one examination question investigation is the 5 th financial knowledge point, the evaluation result is correct, the time is 2min, the request prompt number is 2 times, the corresponding model input is the vector 0,1, 0,1, the term "wherein bits 5, 2 x n+5, and 2 x n+2 x m+5 are 1 and the remaining bits are 0. Correspondingly, [0, 1., 0,. 1, inputting a risk assessment model corresponding to the financial knowledge reserve condition assessment type, input of the corresponding financial knowledge reserve case. In the risk assessment model of the assessment type, 0, 0..1..the..and..the third party is a blood-level-reducing agent, then the vector after the dimension reduction is input into an LSTM model for evaluation, a vector of output 1*N [0.625,0.783..0.918 ], each value in the vector representing a grasp of the knowledge point. Further, after all the evaluation questions of the financial knowledge reserve situation are completed, the evaluation result of the first-level influence factor of the financial knowledge reserve situation is completed. In addition, since the LSTM model can simulate the memory capacity of a person, when a client answers the same knowledge point for a plurality of times, the grasping degree of the knowledge point is gradually increased, and the LSTM model quantifies the interaction between the knowledge points and is embodied in the output 1*N vector.

In some embodiments, the evaluation data of the influence factors of the evaluation type is obtained by digitally signing the acquired evaluation data of the influence factors of the evaluation type at the client, and the evaluation data of the influence factors of the evaluation type is input into a risk evaluation model corresponding to the evaluation type to obtain a risk evaluation result of the client for the evaluation type, including: checking labels are carried out on the evaluation data of the influence factors of the evaluation types; after the verification passes, the evaluation data of the influence factors of the evaluation type are input into a risk evaluation model corresponding to the evaluation type, and a risk evaluation result of the client aiming at the evaluation type is obtained.

The digital signature (also called public key digital signature) is a section of digital string which can not be forged by others only generated by the sender of the information, and is also a valid proof for the authenticity of the information sent by the sender of the information. A set of digital signatures generally defines two complementary operations, one for digital signatures and the other for verifying digital signatures.

Illustratively, as shown in fig. 3, in response to evaluation data of influence factors for each of a plurality of evaluation types submitted by a client through a client 31, the client 31 digitally signs the evaluation data (i.e., test question answer results, answer consuming time, and number of request prompts). Correspondingly, after acquiring the evaluation data of the influence factors of the client for each of the multiple evaluation types, the server 32 performs signature verification on the evaluation data of the influence factors of the evaluation types, and inputs the evaluation data of the influence factors of the evaluation types into the risk evaluation model corresponding to the evaluation type after the signature verification passes, so as to obtain a risk evaluation result of the client for the evaluation type. By digitally signing the evaluation data at the client 31 and signing the received evaluation data at the server 32, tampering of the evaluation data in the transmission process is avoided, the evaluation data is accurate and real, and accuracy of risk evaluation results of the client for the evaluation type is improved.

Further, considering that the storage system in the related art generally uses a conventional database as a persistence layer, but cannot meet the throughput of a large data volume, and cannot complete the fusion of model training and reasoning functions, the conventional database is not suitable for storing training data to be provided to a model. To address this issue, in some embodiments, the evaluation data of the influencing factors of the evaluation type are also recorded in the database and used for updating the risk evaluation model of the corresponding evaluation type after the invalid data is removed.

For example, after the evaluation data of the influence factor of each of the multiple evaluation types is obtained, the evaluation data (i.e. the answer result, the answer consuming time and the request prompting number) of the influence factor of each evaluation type are stored in a first database (for example Hive), the basic information of the client is stored in a second database (for example Mysql), and the data interaction between the first database and the second database is completed by using a timing data transmission task tool (for example sqop). Further, after data collection is completed, data cleaning is performed on the evaluation data in the first database by using a data cleaning tool and a data cleaning algorithm periodically to obtain cleaned evaluation data, and the cleaned evaluation data is stored as training data for updating a risk evaluation model corresponding to an evaluation type, wherein the data cleaning algorithm comprises at least one of abnormal value detection, missing value processing, smoothing processing and filtering processing. Correspondingly, table 3 is a training data model provided in an exemplary embodiment of the present application.

TABLE 3 Table 3

Field name	Field type	Main key	External key	Whether or not it can be empty	Field description
						item_id	int	√	×	NOT NULL	Data numbering
ques_id	Int	×	×	NOT NULL	Test question numbering
						answer	varchar	×	×	NOT NULL	Results of the answers
rel_sec_id	varchar	×	×	NOT NULL	Consider the number of secondary factors
						time_scan	varchar	×	×	NOT NULL	Time spent on answering
tip_count	int	×	×	NOT NULL	Request hint quantity
						model_input	varchar	×	×	NOT NULL	Model input
model_output	varchar	×	×	NOT NULL	Model output

As shown in Table 3, the training data model includes data number, test question number, answer result, investigation secondary factor number, answer time consumption, number of request prompts, model input and model output. Further, each time the evaluation data is collected to a certain amount, the model starts an update strategy, and corresponding model parameters are adjusted through model training. When the risk assessment model of the corresponding assessment type is required to be updated, in order to ensure that the updating of the risk assessment model does not affect the assessment service of the risk assessment model, the risk assessment model can be solved through two sets of models, namely a set of service models and a set of training models, wherein the service models are used for interacting with clients, and the training models are used for model training and updating model parameters based on the assessment data; when model parameters need to be updated, the training model access line is changed into a service model, the service model is offline and is changed into a training model, and model training is performed based on training data.

Illustratively, fig. 6 is a schematic diagram of model training of the financial knowledge reserve situation assessment type provided in an exemplary embodiment of the present application. As shown in fig. 6, where 32 is the number of users, max_len is the maximum number of questions among the 32 users, the number N of secondary influencing factors is 124, the value of the consumed time discrete M is 101, the value of the request prompt number P is 10, the value of the LSTM number in the hidden_layer is 200, and an encoder is provided between the input_layer and the hidden_layer, and the dimension reduction situation is shown in fig. 6; in addition, since the dropout technology is a method for randomly inactivating part of nodes during training of the neural network, the aim is to alleviate the overfitting phenomenon of the model, so that in the model training process, a dropout layer is added behind the hidden_layer for realizing random inactivation, and the dropout layer inactivates part of nodes with preset probability and performs corresponding parameter updating during backward propagation.

Correspondingly, when the first service model based on the LSTM is required to be updated, the first training model based on the LSTM is accessed to be used for interacting with a customer, training data of the financial knowledge reserve condition is input into the first service model based on the LSTM for model training, wherein the first training model based on the LSTM is a risk assessment model which is trained and used for carrying out the financial knowledge reserve condition. When the second service model based on the LSTM is required to be updated, the second training model based on the LSTM is accessed to be used for interacting with clients, training data of the inauguration investment experience is input into the second service model based on the LSTM to carry out model training, wherein the second training model based on the LSTM is a risk assessment model which is trained and used for carrying out inauguration investment experience. When the third service model based on the LSTM is required to be updated, the third training model based on the LSTM is accessed to be used for interacting with clients, and training data expected by future dominant incomes is input into the third service model based on the LSTM for model training, wherein the third training model based on the LSTM is a risk assessment model which is trained for carrying out risk investment experience. When the LSTM-based fourth service model is required to be updated, the LSTM-based fourth training model access line is used for interacting with clients, and the investment attitude training data are input into the LSTM-based fourth service model to perform model training, wherein the LSTM-based fourth training model is a trained risk assessment model for performing the investment attitude.

Fig. 7 is a schematic architecture diagram of an LSTM-based customer financial risk assessment system according to an exemplary embodiment of the present application. As shown in fig. 7, a customer accesses a risk assessment service through a personal internet banking program, reads personal basic information such as asset conditions and professions in a personal information database in an assessment process, reads secondary factors and test question contents through an assessment information database, then the customer answers the test questions, transmits answer results, answer consuming time and request prompting times information to a risk assessment model (service), and transmits the answer results, answer consuming time and request prompting times information to a training data storage database as source data of model training. The risk assessment model (service) is processed to obtain an assessment result, the assessment result is returned to the front-end page, and the client can obtain the financial risk assessment result based on the display on the front-end page. The risk assessment model (training) periodically reads training data in a training data storage database to complete model parameter updating; in addition, the risk assessment model (service) and the risk assessment model (training) are switched periodically, roles of the risk assessment model (service) and the risk assessment model (training) are exchanged after the switching, the risk assessment model (training) is used for outputting financial risk assessment results based on response results, response time consumption and request prompting times of clients, and the current risk assessment model (service) reads training data in a training data storage database to update own parameters.

Fig. 8 is another flow chart of the customer financial risk assessment method based on LSTM according to the exemplary embodiment of the present application. As shown in fig. 8, the client financial risk assessment method based on LSTM in the embodiment of the present application includes the following steps:

s801, entering financial risk assessment.

The client opens a login page of the internet banking service through a browser in the client, inputs a user name and a password to log in, then enters a financial module to select a target financial product, submits a purchase request, correspondingly, the server responds to the purchase request to detect whether the client performs financial risk assessment or whether the current risk assessment result of the client is matched with a risk class portrait of the target financial product, and if the client does not perform financial risk assessment or the current risk assessment result of the client is not matched with the risk class portrait of the target financial product, the evaluation questions of each assessment type are sequentially displayed in the client.

S802, filling in basic information.

Correspondingly, the first part of the test questions requires the customer to fill in the personal basic information, including occupation, annual income and asset status.

S803, evaluating the financial knowledge reserve condition.

The evaluation of the financial knowledge storage condition is carried out, the answer result, the answer consumption time and the request prompting times of each test question are collected, the evaluation data of the influence factors corresponding to the evaluation type are collected, the evaluation data are input into a risk evaluation model of the financial knowledge storage condition, and the evaluation result is a one-dimensional vector.

S804, evaluating the inauguration investment experience.

The method comprises the steps of performing evaluation of the inauguration investment experience, collecting answer results, answer consumption time and request prompting times of each test question, collecting evaluation data of influence factors corresponding to the evaluation type, and inputting the evaluation data into a inauguration evaluation model of the inauguration investment experience, wherein the evaluation result is a one-dimensional vector.

S805, future available income expectation evaluation.

The future available income expectation evaluation is carried out, the answer result, the answer consumption time and the request prompting times of each test question are collected, the evaluation data of the influence factors corresponding to the evaluation type are collected, the evaluation data are input into the future available income expectation risk evaluation model, and the evaluation result is a one-dimensional vector.

S806, evaluating the investment attitude.

The method comprises the steps of carrying out investment attitude assessment, collecting answer results, answer consumption time and request prompting times of each test question, collecting assessment data of influence factors corresponding to the assessment type, and inputting the assessment data into a risk assessment model of the investment attitude, wherein the assessment result is a one-dimensional vector.

S807, determining an evaluation result.

Correspondingly, determining a risk assessment image of the client based on 4 one-dimensional vectors and basic information of the client, comparing the risk assessment image of the client with a risk level image of a target financial product, if the client meets the risk level image of the target financial product, determining that the financial risk assessment result of the client for the target financial product passes, and outputting a purchase page of the target financial product or a target financial product list containing the target financial product; if the client does not meet the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is not passed, and returning to execute the step of inputting the assessment data of the influence factors of the assessment type into the risk assessment model of the corresponding assessment type to obtain the risk assessment result of the client for the assessment type.

In summary, the present application has at least the following advantages:

1. the method has the advantages that the evaluation data of the influence factors of the clients aiming at each evaluation type in a plurality of evaluation types are obtained, the evaluation data are input into the LSTM-based risk evaluation model of the corresponding evaluation type, the influence of the plurality of influence factors in the risk bearing capacity, the interaction among the factors, the response result, the response consumption time and the request prompting times are fully considered, the obtained risk evaluation result of the clients aiming at the evaluation types is more accurate, the deviation of the risk bearing capacity of the clients for purchasing financial products is reduced, and the matching degree of the purchased financial products and the risk bearing capacity of the clients is further improved.

2. Since the memory capacity of the LSTM can simulate the forgetting characteristics of people, the risk bearing capacity of the customer can be reasonably quantified, and the LSTM can also quantify the interaction between different subelements, for example, if the annual rate calculation convenience score of the customer in the financial product is higher, the score in the aspect of the financial redemption rule is also higher, and by introducing the LSTM, the interaction between different subelements can be quantified, which cannot be accomplished by the statistical method in the related art.

3. After each time of evaluation is completed, the method and the system obtain 4 one-dimensional vectors, namely a 4 XN state matrix, can intuitively reflect the risk bearing capacity of the client, can be used for recommending matched financial products for the client, and further provide a certain guarantee for the recommending function expansion of the financial module.

4. The input single-heat codes are subjected to dimension reduction by the encoder introduced into the self-encoder, so that the loss of important information can be reduced to the minimum, and the accuracy of a risk assessment model is further improved.

The following are device embodiments of the present application, which may be used to perform method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

Fig. 9 is a schematic structural diagram of an LSTM-based customer financial risk assessment device according to an exemplary embodiment of the present application. As shown in fig. 9, the LSTM-based customer financial risk assessment apparatus 90 includes an acquisition module 91, an assessment module 92, a first determination module 93, and a second determination module 94, where:

the acquiring module 91 is configured to acquire evaluation data of influence factors of a customer for each of multiple evaluation types, where the multiple evaluation types include financial knowledge storage conditions, inauguration investment experience, future dominant income expectations and investment attitudes, and the evaluation data includes test question answering results, answering consumption time and request prompting times;

the evaluation module 92 is configured to input, for evaluation data of influence factors of each of multiple evaluation types, the evaluation data of the influence factors of the evaluation type into a risk evaluation model of the corresponding evaluation type, to obtain a risk evaluation result of a client for the evaluation type, where the risk evaluation model is a model based on a long-short-term memory network LSTM;

a first determining module 93, configured to determine, according to a risk assessment result of a client for each of a plurality of assessment types and basic information of the client, a risk assessment portrait of the client, where the basic information includes occupation, annual income and asset condition;

The second determining module 94 is configured to determine a financial risk assessment result of the client for the target financial product based on the risk assessment image and the risk level image of the target financial product.

In one possible implementation, the evaluation module 92 may be specifically configured to: performing single-heat coding on the evaluation data of the influence factors of the evaluation type to obtain first coded data; inputting the first encoded data into an encoder for dimension reduction processing to obtain second encoded data, wherein the dimension of the second encoded data is smaller than that of the first encoded data; and inputting the second coded data into the LSTM model for risk assessment, and obtaining a risk assessment result of a customer aiming at an assessment type.

In one possible implementation, the second determining module 94 may be specifically configured to: determining whether the customer meets a risk level representation of the target financial product based on the risk assessment representation; if the client meets the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is passed, and outputting a purchase page of the target financial product or a target financial product list containing the target financial product; if the client does not meet the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is failed, and returning to execute the step of inputting the assessment data of the influence factors of the assessment type into the risk assessment model of the corresponding assessment type to obtain the risk assessment result of the client for the assessment type.

In a possible implementation, the second determining module 94 may also be configured to: determining whether the bearing capacity of the client for each factor corresponding to the evaluation type accords with the risk level portrait of the target financial product; if the bearing capacity of the client for each factor corresponding to the evaluation type accords with the risk level portrait of the target financial product, determining that the client meets the risk level portrait of the target financial product; if the bearing capacity of the client corresponding to any factor of the evaluation type does not meet the risk level portrait of the target financial product, determining that the client does not meet the risk level portrait of the target financial product.

In one possible implementation manner, the evaluation data of the influence factors of the evaluation type are obtained by performing digital signature on the acquired evaluation data of the influence factors of the evaluation type at the client, and the evaluation data of the influence factors of the evaluation type are input into a risk evaluation model corresponding to the evaluation type to obtain a risk evaluation result of the client for the evaluation type, and the method includes: checking labels are carried out on the evaluation data of the influence factors of the evaluation types; after the verification passes, the evaluation data of the influence factors of the evaluation type are input into a risk evaluation model corresponding to the evaluation type, and a risk evaluation result of the client aiming at the evaluation type is obtained.

In a possible implementation manner, the evaluation data of the influence factors of the evaluation type are also recorded in the database, and are used for updating the risk evaluation model corresponding to the evaluation type after invalid data are removed.

The customer financial risk assessment device based on the LSTM provided in this embodiment of the present application may execute the technical solution shown in the foregoing method embodiment, and its implementation principle and beneficial effects are similar, and no redundant description is given here.

It should be noted that, it should be understood that the division of the modules of the above apparatus is merely a division of a logic function, and may be fully or partially integrated into a physical entity or may be physically separated. And these modules may all be implemented in software in the form of calls by the processing element; or can be realized in hardware; the method can also be realized in a form of calling software by a processing element, and the method can be realized in a form of hardware by a part of modules. For example, the processing module may be a processing element that is set up separately, may be implemented in a chip of the above-mentioned apparatus, or may be stored in a memory of the above-mentioned apparatus in the form of program codes, and the functions of the above-mentioned processing module may be called and executed by a processing element of the above-mentioned apparatus. The implementation of the other modules is similar. In addition, all or part of the modules can be integrated together or can be independently implemented. The processing element here may be an integrated circuit with signal processing capabilities. In implementation, each step of the above method or each module above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in a software form.

For example, the modules above may be one or more integrated circuits configured to implement the methods above, such as: one or more application specific integrated circuits (Application Specific Integrated Circuit, abbreviated as ASIC), or one or more microprocessors (Digital Signal Processor, abbreviated as DSP), or one or more field programmable gate arrays (Field Programmable Gate Array, abbreviated as FPGA), or the like. For another example, when a module above is implemented in the form of a processing element scheduler code, the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processor that may invoke the program code. For another example, the modules may be integrated together and implemented in the form of a System-On-a-Chip (SOC).

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the processes or functions in accordance with embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, by wired (e.g., coaxial cable, fiber optic, digital subscriber line (Digital Subscriber Line, simply DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means from one website, computer, server, or data center. Computer readable storage media can be any available media that can be accessed by a computer or data storage devices, such as servers, data centers, etc., that contain an integration of one or more available media. Usable media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital versatile discs (Digital Video Disc, abbreviated to DVD)), or semiconductor media (e.g., solid State Disk (SSD)), etc.

Fig. 10 is a schematic structural diagram of an electronic device according to an exemplary embodiment of the present application. As shown in fig. 10, the electronic apparatus 100 of the present embodiment includes:

at least one processor 101; and a memory 102 communicatively coupled to the at least one processor;

wherein the memory 102 stores instructions executable by the at least one processor 101 to cause the electronic device to perform the method as described in any of the embodiments above.

Alternatively, the memory 102 may be separate or integrated with the processor 101.

The memory 102 may include a high-speed random access memory (Random Access Memory, simply referred to as RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

The processor 101 may be a central processing unit (Central Processing Unit, CPU for short), or an application specific integrated circuit (Application Specific Integrated Circuit, ASIC for short), or one or more integrated circuits configured to implement embodiments of the present application. Specifically, when implementing the customer financial risk assessment method based on LSTM described in the foregoing method embodiment, the electronic device may be, for example, a customer financial risk assessment device based on LSTM deployed in a vehicle, such as a processing chip, or an electronic device with a processing function, such as a server.

Optionally, the electronic device may further comprise a communication interface 103. In a specific implementation, if the communication interface 103, the memory 102, and the processor 101 are implemented independently, the communication interface 103, the memory 102, and the processor 101 may be connected to each other and perform communication with each other through buses. The bus may be an industry standard architecture (Industry Standard Architecture, ISA) bus, an external device interconnect (Peripheral Component, PCI) bus, or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, among others. Buses may be divided into address buses, data buses, control buses, etc., but do not represent only one bus or one type of bus.

Alternatively, in a specific implementation, if the communication interface 103, the memory 102, and the processor 101 are integrated on a chip, the communication interface 103, the memory 102, and the processor 101 may complete communication through internal interfaces.

The implementation principle and technical effects of the electronic device provided in this embodiment may be referred to the foregoing embodiments, and will not be described herein again.

The embodiment of the present application further provides a computer readable storage medium, where computer execution instructions are stored, where the computer execution instructions are used to implement the method steps in the method embodiment described above when executed, and specific implementation manner and technical effect are similar, and are not repeated herein.

The computer readable storage medium may be implemented by any type or combination of volatile or nonvolatile Memory devices, such as static random access Memory (Static Random Access Memory, SRAM for short), electrically erasable programmable Read-Only Memory (Electrically Erasable Programmable Read Only Memory, EEPROM for short), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM for short), programmable Read-Only Memory (Programmable Read Only Memory, PROM for short), read Only Memory (ROM for short), magnetic Memory, flash Memory, magnetic disk, or optical disk. A readable storage medium can be any available medium that can be accessed by a general purpose or special purpose computer.

An exemplary readable storage medium is coupled to the processor such the processor can read information from, and write information to, the readable storage medium. In the alternative, the readable storage medium may be integral to the processor. The processor and the readable storage medium may reside in an application specific integrated circuit. Of course, the processor and the readable storage medium may also reside as discrete components in an LSTM-based customer financial risk assessment device.

The embodiment of the present application further provides a computer program product, which includes a computer program, and when the computer program is executed, performs the method steps in the embodiment of the method, and the specific implementation manner and the technical effect are similar, and are not repeated herein.

Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the application following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the application pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.

It is to be understood that the present application is not limited to the precise arrangements and instrumentalities shown in the drawings, which have been described above, and that various modifications and changes may be effected without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims

1. The LSTM-based client financial risk assessment method is characterized by comprising the following steps of:

acquiring evaluation data of influence factors of a customer aiming at each of a plurality of evaluation types, wherein the plurality of evaluation types comprise financial knowledge storage conditions, inauguration investment experience, future dominant income expectations and investment attitudes, and the evaluation data comprise test question answering results, answering consumption time and request prompting times;

Inputting the evaluation data of the influence factors of the evaluation types into a risk evaluation model of the corresponding evaluation type aiming at the evaluation data of the influence factors of each evaluation type in the plurality of evaluation types to obtain a risk evaluation result of the client aiming at the evaluation type, wherein the risk evaluation model is a model based on a long-short-term memory network LSTM;

determining a risk assessment portrait of the client according to a risk assessment result of the client for each of the multiple assessment types and basic information of the client, wherein the basic information comprises occupation, annual income and asset conditions;

and determining a financial risk assessment result of the client aiming at the target financial product based on the risk assessment image and the risk level image of the target financial product.

2. The LSTM-based risk assessment method for customer financial management according to claim 1, wherein the risk assessment model includes an encoder of a self-encoder and an LSTM model, the inputting the assessment data of the influence factors of the assessment type into the risk assessment model of the corresponding assessment type, to obtain the risk assessment result of the customer for the assessment type, includes:

Performing single-heat coding on the evaluation data of the influence factors of the evaluation types to obtain first coded data;

inputting the first encoded data to the encoder for dimension reduction processing to obtain second encoded data, wherein the dimension of the second encoded data is smaller than that of the first encoded data;

and inputting the second coded data into the LSTM model for risk assessment, and obtaining a risk assessment result of the client aiming at the assessment type.

3. The LSTM-based financial risk assessment method according to claim 1 or 2, wherein the determining, based on the risk assessment image and a risk level portrait of a target financial product, a financial risk assessment result of the client for the target financial product includes:

determining whether the customer meets a risk level representation of the target financial product based on the risk assessment representation;

if the client meets the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is passed, and outputting a purchase page of the target financial product or a target financial product list containing the target financial product;

If the client does not meet the risk level portrait of the target financial product, determining that the financial risk assessment result of the client for the target financial product is failed, and returning to execute the step of inputting the assessment data of the influence factors of the assessment type into the risk assessment model of the corresponding assessment type to obtain the risk assessment result of the client for the assessment type.

4. The LSTM based customer financial risk assessment method according to claim 3, wherein the risk assessment image includes bearing capacity of the customer for each factor corresponding to the assessment type, and the determining whether the customer meets the risk level image of the target financial product based on the risk assessment image includes:

determining whether the bearing capacity of the client for each factor corresponding to the evaluation type accords with the risk level portrait of the target financial product;

if the bearing capacity of the client corresponding to each factor of the evaluation type accords with the risk level portrait of the target financial product, determining that the client meets the risk level portrait of the target financial product;

And if the bearing capacity of the client corresponding to any factor of the evaluation type does not accord with the risk level portrait of the target financial product, determining that the client does not meet the risk level portrait of the target financial product.

5. The LSTM-based financial risk assessment method according to claim 1 or 2, wherein the assessment data of the impact factors of the assessment type is obtained by digitally signing the collected assessment data of the impact factors of the assessment type at the client, and the inputting the assessment data of the impact factors of the assessment type into a risk assessment model of a corresponding assessment type, to obtain a risk assessment result of the client for the assessment type, includes:

checking labels are carried out on the evaluation data of the influence factors of the evaluation types;

after the verification passes, inputting the evaluation data of the influence factors of the evaluation type into a risk evaluation model corresponding to the evaluation type, and obtaining a risk evaluation result of the client for the evaluation type.

6. The LSTM based customer financial risk assessment method according to claim 1 or 2, wherein the assessment data is obtained by:

After the customer selects the target financial product through man-machine interaction, if the customer is determined to not meet the risk level portrait of the target financial product according to the current risk assessment result of the customer, the assessment questions of each assessment type in the multiple assessment types are displayed so as to acquire the assessment data of the influence factors corresponding to each assessment type.

7. The LSTM-based customer financial risk assessment method according to claim 6, wherein the assessment data of the influencing factors of the assessment type is further recorded in a database, and is used for updating the risk assessment model of the corresponding assessment type after invalid data are removed.

8. Customer financial risk assessment device based on LSTM, characterized in that includes:

the system comprises an acquisition module, a judgment module and a judgment module, wherein the acquisition module is used for acquiring evaluation data of influence factors of a client aiming at each evaluation type in a plurality of evaluation types, the plurality of evaluation types comprise financial knowledge storage conditions, inauguration investment experience, future dominant income expectations and investment attitudes, and the evaluation data comprise test question answering results, answering consumption time and request prompting times;

the evaluation module is used for inputting the evaluation data of the influence factors of the evaluation types into a risk evaluation model of the corresponding evaluation type aiming at the evaluation data of the influence factors of each evaluation type in the plurality of evaluation types to obtain a risk evaluation result of the client aiming at the evaluation type, wherein the risk evaluation model is a model based on a long-short-term memory network LSTM;

The first determining module is used for determining the risk assessment portrait of the client according to the risk assessment result of the client for each assessment type in the plurality of assessment types and the basic information of the client, wherein the basic information comprises occupation, annual income and asset condition;

and the second determining module is used for determining financial risk assessment results of the clients aiming at the target financial products based on the risk assessment image and the risk level image of the target financial products.

9. An electronic device, comprising: a processor, and a memory communicatively coupled to the processor;

the memory is used for storing computer execution instructions;

the processor is configured to execute the computer-executable instructions to implement the LSTM-based customer financial risk assessment method according to any one of claims 1 to 7.

10. A computer readable storage medium having stored therein computer executable instructions which when executed are adapted to implement the LSTM based customer financial risk assessment method of any one of claims 1 to 7.