CN111681091A

CN111681091A - Financial risk prediction method and device based on time domain information and storage medium

Info

Publication number: CN111681091A
Application number: CN202010804538.3A
Authority: CN
Inventors: 蓝利君; 孙艺芙; 郭清宇; 李超; 王翔
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-08-12
Filing date: 2020-08-12
Publication date: 2020-09-18
Anticipated expiration: 2040-08-12
Also published as: CN111681091B

Abstract

The invention provides a financial risk prediction method based on time domain information, a device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring unlabeled financial information sample characteristics matched with a target user; the method comprises the steps of determining different financial service scenes corresponding to unlabeled financial information sample characteristics matched with a target user respectively, calling financial information processing models trained through time domain training samples in an antagonistic manner, and performing risk prediction processing on unlabeled sample characteristic sets through the financial information processing models to obtain financial risk prediction results of the target user.

Description

Financial risk prediction method and device based on time domain information and storage medium

Technical Field

The present invention relates to data processing technology in neural network models, and in particular, to a method and an apparatus for predicting financial risk based on time domain information, an electronic device, and a storage medium.

Background

The recognition of each category based on deep learning is always an important tool for solving a large amount of data points in each application scene. For example, in application scenarios such as images, natural language processing, risk early warning, and the like, large-scale classification and identification of a large amount of data are realized, so that relevant classification prediction results are rapidly and accurately obtained, and the function implementation of the application scenario in which the data are located is accelerated. In risk control in financial business links such as payment, loan, financing, etc., data distribution of different customer groups is widely different, and there are numerous customer groups with small sample characteristics. Therefore, the traditional model screening method depends on a data set which is large in data volume, long in time span and rich in labeled samples, but in the actual production of the financial wind control scene, the use of the model faces the data problem that the data volume is small and positive and negative samples are quite unbalanced, so that the model overfitting phenomenon easily occurs, and the accuracy of the prediction result of the financial information processing model is reduced along with the increase of the use time.

Disclosure of Invention

In view of this, embodiments of the present invention provide a financial risk prediction method and apparatus based on time domain information, an electronic device, and a storage medium, which can predict a risk of a target user in a financial scenario through a financial information processing model, reduce complexity of financial information processing, and make the generalization capability and data processing capability of the financial information processing model stronger while considering training accuracy, so as to adapt to different data processing environments and reduce robustness.

The technical scheme of the embodiment of the invention is realized as follows:

the embodiment of the invention also provides a financial risk prediction method based on the time domain information, which comprises the following steps:

responding to the financial risk prediction request, and acquiring unlabeled financial information sample characteristics matched with the target user;

determining different financial service scenes respectively corresponding to the unlabeled financial information sample characteristics matched with the target user;

determining dynamic noise matched with the different financial business scenes based on the different financial business scenes;

based on the dynamic noise, dynamically denoising the unlabeled financial information sample characteristics matched with the target user to form an unlabeled sample characteristic set matched with the corresponding financial business scene;

calling a financial information processing model for countertraining through a time domain training sample, and performing risk prediction processing on the unlabeled sample feature set through the financial information processing model to obtain a financial risk prediction result of the target user;

and outputting the financial risk prediction result of the target user.

In the above scheme, the method further comprises:

acquiring characteristics of a target user set and historical parameters of a financial scene;

obtaining a sample feature set matched with the financial information processing model according to the features of the target user set and historical parameters of financial scenes, wherein the sample feature set comprises sample features of different time domains;

processing the sample feature set based on different time window lengths, and determining a training sample set of different time domains matched with the financial information processing model, wherein the training sample set comprises at least one group of training samples;

training the financial information processing model according to the training sample sets of different time domains matched with the financial information processing model, and determining model parameters matched with the financial information processing model so as to predict the risk of the target user in the financial scene through the financial information processing model.

In the above scheme, the method further comprises:

determining a dynamic noise threshold value matched with the use environment of the financial information processing model;

denoising the sample feature set according to the dynamic noise threshold value to form a sample feature set matched with the dynamic noise threshold value; alternatively, the first and second electrodes may be,

and determining a fixed noise threshold corresponding to the financial information processing model, and denoising the sample feature set according to the fixed noise threshold to form a sample feature set matched with the fixed noise threshold.

In the foregoing solution, the processing, by the feature extractor network, the training sample set in response to the initial parameter of the feature extractor network to determine an update parameter of the feature extractor network includes:

substituting different training samples in the training sample set into a loss function corresponding to the feature extractor network;

and determining that the feature extractor network corresponds to the update parameter when the loss function meets the corresponding convergence condition.

In the above scheme, the method further comprises:

and sending the target user identification, the model parameters of the financial information processing model and the financial scene identification to a block chain network, so that the nodes of the block chain network fill the target user identification, the model parameters of the financial information processing model and the financial scene identification into a new block, and when the new block is identified in a consistent manner, adding the new block to the tail part of the block chain.

In the above scheme, the method further comprises:

receiving data synchronization requests of other nodes in the blockchain network;

responding to the data synchronization request, and verifying the authority of the other nodes;

and when the authority of the other nodes passes the verification, controlling the current node and the other nodes to carry out data synchronization so as to realize that the other nodes acquire the target user identification, the model parameters of the financial information processing model and the financial scene identification.

In the above scheme, the method further comprises:

responding to a query request, and analyzing the query request to obtain a corresponding object identifier;

acquiring authority information in a target block in a block chain network according to the object identifier;

checking the matching of the authority information and the object identification;

when the authority information is matched with the object identification, acquiring a corresponding target user identification, a model parameter of a financial information processing model and a financial scene identification in the block chain network;

and responding to the query instruction, pushing the acquired corresponding financial information and the target object matched with the basic object to a corresponding client so as to enable the client to acquire the corresponding financial information stored in the blockchain network and the target object matched with the basic object.

The embodiment of the invention also provides a financial risk prediction device based on the time domain information, which comprises:

the first information transmission module is used for responding to the financial risk prediction request and acquiring the unlabeled financial information sample characteristics matched with the target user;

the information processing module is used for determining different financial service scenes corresponding to the non-label financial information sample characteristics matched with the target user respectively;

the information processing module is used for determining dynamic noise matched with the different financial business scenes based on the different financial business scenes; based on the dynamic noise, dynamically denoising the unlabeled financial information sample characteristics matched with the target user to form an unlabeled sample characteristic set matched with the corresponding financial business scene;

the information processing module is used for calling a financial information processing model for countertraining through a time domain training sample, and performing risk prediction processing on the unlabeled sample feature set through the financial information processing model to obtain a financial risk prediction result of the target user;

and the information processing module is used for outputting the financial risk prediction result of the target user.

In the above scheme, the second information transmission module is configured to obtain characteristics of a target user set and historical parameters of a financial scenario;

the training module is used for determining different time window lengths, processing the sample feature set according to the corresponding time window lengths and determining training sample sets of different time domains matched with the financial information processing model, wherein the training sample sets comprise at least one group of training samples;

the training module is used for training the financial information processing model according to the training sample sets of different time domains matched with the financial information processing model, and determining model parameters matched with the financial information processing model so as to predict the risk of the target user in the financial scene through the financial information processing model.

In the above scheme, the training module is configured to determine a dynamic noise threshold that matches a usage environment of the financial information processing model;

In the above scheme, the training module is configured to determine time windows of different lengths based on a usage environment of the financial information processing model;

the training module is used for determining timestamp parameters carried by different sample data in the sample feature set;

and the training module is used for processing the timestamp parameters carried by different sample data in the sample characteristic set according to a random gradient descent algorithm matched with the financial information processing model and time windows with different lengths, and determining training sample sets of different time domains matched with the financial information processing model.

In the above scheme, the training module is configured to train the financial information processing model according to the training sample set, and determine a model parameter of a feature extractor network in the financial information processing model;

the training module is used for training the financial information processing model according to the training sample set and determining model parameters of a sample classifier network in the financial information processing model;

and the training module is used for training the financial information processing model according to the training sample set and determining the model parameters of the domain classifier network in the financial information processing model.

In the above scheme, the training module is configured to process the training sample set through a feature extractor network in the financial information processing model to determine initial parameters of the feature extractor network;

the training module is used for responding to initial parameters of the feature extractor network, processing the training sample set through the feature extractor network and determining updating parameters of the feature extractor network;

and the training module is used for carrying out iterative updating on the parameters of the feature extractor network through the training sample set according to the updated parameters of the feature extractor network so as to extract the feature embedded vector of each sample in the training sample set.

In the above scheme, the training module is configured to substitute different training samples in the training sample set into a loss function corresponding to the feature extractor network;

the training module is configured to determine that the feature extractor network corresponds to the update parameter when the loss function satisfies a corresponding convergence condition.

In the above scheme, the training module is configured to determine a loss function corresponding to the feature extractor network;

the training module is used for carrying out iterative updating on the parameters of the feature extractor network according to the updated parameters of the feature extractor network; until the loss function of the feature extractor network reaches a corresponding convergence condition, and based on the parameters in the feature extractor network, a feature embedding vector of each sample in the training sample set can be extracted.

In the above scheme, the training module is configured to process the training sample set through a sample classifier network in the financial information processing model to determine initial parameters of the sample classifier network;

the training module is used for responding to initial parameters of the sample classifier network, processing the training sample set through the sample classifier network and determining updating parameters of the sample classifier network;

and the training module is used for performing iterative updating on the parameters of the sample classifier network through the training sample set according to the updated parameters of the sample classifier network so as to determine risk prediction results of different samples based on corresponding sample labels and the feature embedded vectors of each sample.

In the above scheme, the training module is configured to substitute different training samples in the training sample set into a loss function corresponding to the sample classifier network;

the training module is configured to determine that the sample classifier network corresponds to the update parameter when the loss function satisfies a corresponding convergence condition.

In the above scheme, the training module is configured to determine a loss function corresponding to the sample classifier network;

the training module is used for carrying out iterative updating on the parameters of the sample classifier network according to the updated parameters of the sample classifier network; until the loss function of the sample classifier network reaches the corresponding convergence condition, and based on the corresponding sample label and the feature embedding vector of each sample, determining the risk prediction results of different samples.

In the foregoing solution, the training module is configured to respond to an initial parameter of the domain classifier network, process the training sample set through the domain classifier network, and determine an update parameter of the domain classifier network;

and the training module is used for carrying out iterative updating on the parameters of the domain classifier network through the training sample set according to the updating parameters of the domain classifier network so as to judge the time domains corresponding to different samples in the sample set respectively.

In the above scheme, the training module is configured to substitute different training samples in the training sample set into a loss function corresponding to the domain classifier network;

the training module is configured to determine that the domain classifier network corresponds to the update parameter when the loss function satisfies a corresponding convergence condition.

In the above scheme, the training module is configured to determine a loss function corresponding to the domain classifier network;

the training module is used for carrying out iterative updating on the parameters of the domain classifier network according to the updated parameters of the domain classifier network; until the loss function of the domain classifier network reaches the corresponding convergence condition, and based on the parameters in the domain classifier network, the time domains corresponding to different samples in the sample set can be judged.

In the above scheme, the information processing module is configured to send the target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier to a blockchain network, so that a node of the blockchain network fills the target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier into a new block, and when the new block is identified consistently, the new block is added to the tail of the blockchain.

In the above solution, the information processing module is configured to receive a data synchronization request from another node in the block chain network;

the information processing module is used for responding to the data synchronization request and verifying the authority of the other nodes;

and the information processing module is used for controlling the current node and the other nodes to carry out data synchronization when the authority of the other nodes passes verification so as to realize that the other nodes acquire target user identification, model parameters of a financial information processing model and financial scene identification.

In the above scheme, the information processing module is configured to respond to a query request, and parse the query request to obtain a corresponding object identifier;

the information processing module is used for acquiring authority information in a target block in a block chain network according to the object identifier;

the information processing module is used for verifying the matching of the authority information and the object identification;

the information processing module is used for acquiring corresponding target user identification, model parameters of a financial information processing model and financial scene identification in the block chain network when the authority information is matched with the object identification;

and the information processing module is used for responding to the query instruction, and pushing the acquired corresponding financial information and the target object matched with the basic object to the corresponding client so as to realize that the client acquires the corresponding financial information stored in the blockchain network and the target object matched with the basic object.

An embodiment of the present invention further provides an electronic device, where the electronic device includes:

a memory for storing executable instructions;

and the processor is used for implementing the financial risk prediction method based on the time domain information or the financial risk prediction method based on the time domain information when the executable instructions stored in the memory are executed.

The embodiment of the invention has the following beneficial effects:

the method comprises the steps of responding to a financial risk prediction request, and obtaining non-label financial information sample characteristics matched with a target user; determining different financial service scenes respectively corresponding to the unlabeled financial information sample characteristics matched with the target user; determining dynamic noise matched with the different financial business scenes based on the different financial business scenes; based on the dynamic noise, dynamically denoising the unlabeled financial information sample characteristics matched with the target user to form an unlabeled sample characteristic set matched with the corresponding financial business scene; processing the non-labeled financial information sample characteristics through a financial information processing model to form a risk prediction result of the target user; the risk prediction result of the target user is output, so that the risk of the target user in a financial scene can be predicted through the financial information processing model, the complexity of financial information processing is reduced, the generalization capability and the data processing capability of the financial information processing model are stronger while the training accuracy is considered, the method is suitable for different financial data processing environments, and the robustness is reduced.

Drawings

FIG. 1 is a schematic diagram of an environment for using a financial risk prediction method based on time domain information according to an embodiment of the present invention;

fig. 2 is a schematic structural diagram of an electronic device according to an embodiment of the present invention;

FIG. 3 is a schematic flow chart of an alternative method for predicting financial risk based on time domain information according to the present application;

FIG. 4 is a schematic diagram of an alternative model training scheme in accordance with an embodiment of the present invention;

FIG. 5 is a schematic flow chart illustrating an alternative method for predicting financial risk based on time domain information according to the present application;

FIG. 6 is a schematic diagram of an architecture of a financial risk prediction apparatus 100 based on time domain information according to an embodiment of the present invention;

fig. 7 is a schematic structural diagram of a block chain in the block chain network 200 according to an embodiment of the present invention;

fig. 8 is a functional architecture diagram of a blockchain network 200 according to an embodiment of the present invention;

FIG. 9 is a schematic diagram of a front-end display of a method for financial risk prediction based on time domain information provided herein;

FIG. 10 is a schematic diagram illustrating a process of using the method for predicting financial risk based on time domain information provided herein;

fig. 11 is a schematic front-end display diagram of a financial risk prediction method based on time domain information according to the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail with reference to the accompanying drawings, the described embodiments should not be construed as limiting the present invention, and all other embodiments obtained by a person of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

Before further detailed description of the embodiments of the present invention, terms and expressions mentioned in the embodiments of the present invention are explained, and the terms and expressions mentioned in the embodiments of the present invention are applied to the following explanations.

1) In response to the condition or state on which the performed operation depends, one or more of the performed operations may be in real-time or may have a set delay when the dependent condition or state is satisfied; there is no restriction on the order of execution of the operations performed unless otherwise specified.

2) Based on the condition or state on which the operation to be performed depends, when the condition or state on which the operation depends is satisfied, the operation or operations to be performed may be in real time or may have a set delay; there is no restriction on the order of execution of the operations performed unless otherwise specified.

3) Convolutional Neural Networks (CNN Convolutional Neural Networks) are a class of Feed forward Neural Networks (Feed forward Neural Networks) that contain convolution computations and have a deep structure, and are one of the representative algorithms for deep learning (deep). The convolutional neural network has a representation learning (representation learning) capability, and can perform shift-invariant classification (shift-invariant classification) on input information according to a hierarchical structure of the convolutional neural network.

4) And (4) model training, namely performing multi-classification learning on the image data set. The model can be constructed by adopting deep learning frames such as Tensor Flow, torch and the like, and a multi-classification model is formed by combining multiple layers of neural network layers such as CNN and the like. The input of the model is a three-channel or original channel matrix formed by reading an image through openCV and other tools, the output of the model is multi-classification probability, and the webpage category is finally output through softmax and other algorithms. During training, the model approaches to a correct trend through an objective function such as cross entropy and the like.

5) Neural Networks (NN): an Artificial Neural Network (ANN), referred to as Neural Network or Neural Network for short, is a mathematical model or computational model that imitates the structure and function of biological Neural Network (central nervous system of animals, especially brain) in the field of machine learning and cognitive science, and is used for estimating or approximating functions.

6) Transactions (transactions), equivalent to the computer term "Transaction," include operations that need to be committed to a blockchain network for execution and do not refer solely to transactions in the context of commerce, which embodiments of the present invention follow in view of the convention colloquially used in blockchain technology.

For example, a deployment (deployment) transaction is used to install a specified smart contract to a node in a blockchain network and is ready to be invoked; the Invoke (Invoke) transaction is used to append records of the transaction in the blockchain by invoking the smart contract and to perform operations on the state database of the blockchain, including update operations (including adding, deleting, and modifying key-value pairs in the state database) and query operations (i.e., querying key-value pairs in the state database).

7) A Block chain (Block chain) is an encrypted, chained transaction storage structure formed of blocks (blocks).

For example, the header of each block may include hash values of all transactions in the block, and also include hash values of all transactions in the previous block, so as to achieve tamper resistance and forgery resistance of the transactions in the block based on the hash values; newly generated transactions, after being filled into the tiles and passing through the consensus of nodes in the blockchain network, are appended to the end of the blockchain to form a chain growth.

8) A Block chain Network (Block chain Network) incorporates new blocks into a set of nodes of a Block chain in a consensus manner.

9) Ledger (legger) is a general term for blockchains (also called Ledger data) and state databases synchronized with blockchains.

Wherein, the blockchain records the transaction in the form of a file in a file system; the state database records the transactions in the blockchain in the form of different types of Key (Key) Value pairs for supporting fast query of the transactions in the blockchain.

10) Intelligent Contracts (Smart Contracts), also known as Chain codes (Chain codes) or application codes, are programs deployed in nodes of a blockchain network, which execute intelligent Contracts called in received transactions to perform operations of updating or querying key-value data of the account database.

11) Consensus (Consensus), a process in a blockchain network, is used to agree on transactions in blocks among the nodes involved, the agreed blocks are to be appended to the end of the blockchain, and the mechanisms to achieve Consensus include Proof of workload (Po W), Proof of rights and interests (PoS, Proof of stamp), Proof of equity authorization (DPo S, released Proof of stamp), Proof of Elapsed Time (Po ET, Proof of Elapsed Time), etc.

12) K-S test, based on the cumulative distribution function, is used to check whether the two empirical distributions are different or whether one empirical distribution is different from the other ideal distribution. Under the financial wind control scene, the KS value is commonly used as an evaluation index for distinguishing the separation degree of positive and negative samples by the model.

13) The OOT set, i.e., the cross-time validation set isolated in time from the training data, is partitioned from the entire marked dataset using a certain time node as a standard. The time node is preceded by a training test set and followed by an OOT set, and a sample of the latest application date is usually reserved as the OOT to measure the stability of the model in time.

Fig. 1 is a schematic view of a usage scenario of a financial risk prediction method based on time domain information according to an embodiment of the present invention, referring to fig. 1, a terminal (including a terminal 10-1 and a terminal 10-2) is provided with a client capable of displaying software of corresponding financial information, for example, a client or a plug-in for performing financial activities on virtual resources or physical resources or paying through virtual resources (bitcoin or Q coin), a user can obtain and display financial information through the corresponding client, and a corresponding financial information processing process (for example, a process of paying through a WeChat financial payment or lending through a WeChat applet) is triggered in a financial information processing process; the terminal is connected to the server 200 through a network 300, and the network 300 may be a wide area network or a local area network, or a combination of the two, and uses a wireless link to realize data transmission.

As an example, the server 200 is configured to deploy the time domain information-based financial risk prediction apparatus to implement the time domain information-based financial risk prediction method provided by the present invention, so as to obtain unlabeled financial information sample characteristics matching the target user in response to a financial risk prediction request; calling a financial information processing model for countertraining through a time domain training sample, and performing risk prediction processing on the unlabeled sample feature set through the financial information processing model to obtain a financial risk prediction result of the target user; and outputting the financial risk prediction result of the target user.

Of course, before the financial data is processed by the financial information processing model to generate the corresponding prediction result, the financial information processing model needs to be trained, which specifically includes:

acquiring characteristics of a target user set and historical parameters of a financial scene; obtaining a sample feature set matched with the financial information processing model according to the features of the target user set and historical parameters of financial scenes, wherein the sample feature set comprises sample features of different time domains; processing the sample feature set based on different time window lengths, and determining a training sample set of different time domains matched with the financial information processing model, wherein the training sample set comprises at least one group of training samples; training the financial information processing model according to the training sample sets of different time domains matched with the financial information processing model, and determining model parameters matched with the financial information processing model so as to predict the risk of the target user in the financial scene through the financial information processing model.

Of course, the financial risk prediction apparatus based on time domain information provided by the present invention may be applied to a usage environment in which a virtual resource or an entity resource performs financial activities or performs information interaction through an entity financial resource payment environment (including but not limited to various types of entity financial resource change environments) or social software, and financial information of different data sources is usually processed in performing financial activities on various types of entity financial resources or performing payment through a virtual resource, and finally, financial information corresponding to a target object selected by the target User is presented on a User Interface (UI). The financial information (such as user risk judgment) obtained by the user in the current display interface can be called by other application programs.

As described in detail below, the financial risk prediction apparatus based on time domain information according to an embodiment of the present invention may be implemented in various forms, such as a dedicated terminal having a processing function of the financial risk prediction apparatus based on time domain information, or a server having a processing function of the financial risk prediction apparatus based on time domain information, for example, the server 200 in fig. 1. Fig. 2 is a schematic diagram of a composition structure of an electronic device according to an embodiment of the present invention, and it is to be understood that fig. 2 only shows an exemplary structure of a financial risk prediction apparatus based on time domain information, and not a whole structure, and a part of or the whole structure shown in fig. 2 may be implemented as needed.

The financial risk prediction device based on the time domain information provided by the embodiment of the invention comprises: at least one processor 201, memory 202, user interface 203, and at least one network interface 204. The various components in the time domain information based financial risk prediction apparatus are coupled together by a bus system 205. It will be appreciated that the bus system 205 is used to enable communications among the components. The bus system 205 includes a power bus, a control bus, and a status signal bus in addition to a data bus. For clarity of illustration, however, the various buses are labeled as bus system 205 in fig. 2.

The user interface 203 may include, among other things, a display, a keyboard, a mouse, a trackball, a click wheel, a key, a button, a touch pad, or a touch screen.

It will be appreciated that the memory 202 can be either volatile memory or nonvolatile memory, and can include both volatile and nonvolatile memory. The memory 202 in embodiments of the present invention is capable of storing data to support operation of the terminal (e.g., 10-1). Examples of such data include: any computer program, such as an operating system and application programs, for operating on a terminal (e.g., 10-1). The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, and the like, and is used for implementing various basic services and processing hardware-based tasks. The application program may include various application programs.

In some embodiments, the financial risk prediction apparatus based on time domain information provided by the embodiments of the present invention may be implemented by a combination of hardware and software, and as an example, the financial risk prediction apparatus based on time domain information provided by the embodiments of the present invention may be a processor in the form of a hardware decoding processor, which is programmed to execute the financial risk prediction method based on time domain information provided by the embodiments of the present invention. For example, a processor in the form of a hardware decoding processor may employ one or more Application Specific Integrated Circuits (ASICs), DSPs, Programmable Logic Devices (PLDs), Complex Programmable Logic Devices (CPLDs), Field Programmable Gate Arrays (FPGAs), or other electronic components.

As an example of the implementation of the financial risk prediction apparatus based on time domain information provided by the embodiment of the present invention by using a combination of software and hardware, the financial risk prediction apparatus based on time domain information provided by the embodiment of the present invention may be directly embodied as a combination of software modules executed by the processor 201, where the software modules may be located in a storage medium, the storage medium is located in the memory 202, and the processor 201 reads executable instructions included in the software modules in the memory 202, and completes the financial risk prediction method based on time domain information provided by the embodiment of the present invention in combination with necessary hardware (for example, including the processor 201 and other components connected to the bus 205).

By way of example, the Processor 201 may be an integrated circuit chip having Signal processing capabilities, such as a general purpose Processor, a Digital Signal Processor (DSP), or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or the like, wherein the general purpose Processor may be a microprocessor or any conventional Processor or the like.

As an example of the time domain information based financial risk prediction apparatus provided by the embodiment of the present invention implemented by hardware, the apparatus provided by the embodiment of the present invention may be implemented by directly using the processor 201 in the form of a hardware decoding processor, for example, by one or more Application Specific Integrated Circuits (ASICs), DSPs, Programmable Logic Devices (PLDs), Complex Programmable Logic Devices (CPLDs), Field Programmable Gate Arrays (FPGAs), or other electronic components, to implement the time domain information based financial risk prediction method provided by the embodiment of the present invention.

Memory 202 in embodiments of the present invention is used to store various types of data to support the operation of a financial risk prediction device based on time domain information. Examples of such data include: any executable instructions for operating on a time domain information based financial risk prediction apparatus, such as executable instructions, may be embodied in a program for implementing a method for predicting financial risk from time domain information based financial risk according to an embodiment of the present invention.

In other embodiments, the financial risk prediction apparatus based on time domain information provided by the embodiment of the present invention may be implemented in software, and fig. 2 illustrates the financial risk prediction apparatus based on time domain information stored in the memory 202, which may be software in the form of programs and plug-ins, and includes a series of modules, as an example of the programs stored in the memory 202, which may include the financial risk prediction apparatus based on time domain information, and the financial risk prediction apparatus based on time domain information includes the following software modules: a second information transmission module 2081, a training module 2082, a first information transmission module 2083, and an information processing module 2084. When the software modules in the financial risk prediction device based on the time domain information are read into the RAM by the processor 201 and executed, the financial risk prediction method based on the time domain information provided by the embodiment of the invention is implemented, wherein the functions of each software module in the financial risk prediction device based on the time domain information include:

the second information transmission module 2081, configured to obtain characteristics of the target user set and historical parameters of the financial scenario;

the training module 2082 is configured to process the sample feature set based on different time window lengths, and determine a training sample set of different time domains matched with the financial information processing model, where the training sample set includes at least one set of training samples;

the training module 2082 is configured to train the financial information processing model according to training sample sets of different time domains matched with the financial information processing model, and determine model parameters adapted to the financial information processing model, so as to predict risks of a target user in the financial scene through the financial information processing model.

The first information transmission module 2083, configured to respond to the financial risk prediction request, and obtain a non-tag financial information sample characteristic matched with the target user;

the information processing module 2084 is used for processing through a financial information processing model based on the non-tag financial information sample characteristics to form a risk prediction result of the target user;

the information processing module 2084 is configured to output a risk prediction result of the target user.

Before introducing the financial risk prediction method based on time domain information provided by the application, a financial information processing model in a financial wind control scene in the prior art is preferentially described, wherein the conventional machine learning focuses on how to learn a stable and accurate prediction mathematical model from mass data. However, no effective solution is provided for the problem of model decay with time. Most traditional machine learning relies on passive selection of a model with strong generalization performance in the time dimension on an OOT verification set to reduce the risk of model effect decay over time. And training the model by modes of mode-entering feature selection, hyper-parameter adjustment and the like, and selecting the model with the minimum attenuation compared with the training set effect on the OOT verification set. However, in a financial wind control scene, data distribution differences of different customer groups are large, sufficient labeled data are often difficult to collect, and the prior art is not suitable for prediction model training in the scene, specifically, the defects in the prior art are mainly focused on 1) a model screening method of an OOT verification set depends on a data set which is large in data size, long in time span and rich in labeled samples. The OOT set is a cross-time verification set isolated from training data in time, and is divided from the whole marked data set by taking a certain time node as a standard. The time node is preceded by a training test set and followed by an OOT set, and a sample of the latest application date is usually reserved as the OOT to measure the stability of the model in time. Because the OOT verification set is divided from the whole marked data set by taking a certain time node as a reference, the labeled data of all OOT verification sets cannot be used for model training, and therefore under the condition that the labeled data set is limited, the OOT verification set is too large, the model training data is insufficient, and the accuracy is reduced. If the OOT verification set is too small, the model screening effect is not good, and the model which is stable over time cannot be screened effectively. However, in the actual production of the financial wind control scene, the use of the model faces the data problems that the data volume is small and the positive and negative samples are not balanced. Too big OOT verifies that the set can make model learning insufficient, not only can't select stable accurate model but also can produce negative effects to the precision of model prediction even. 2) The acquisition cost for acquiring the labeled samples in the financial environment is very high, the unlabeled samples with long time span and abundant sample amount exist, and if only the labeled samples with high cost are used for model screening, the cost for system construction and maintenance is increased, and all data cannot be effectively utilized. 3) The selection of the OOT time node also affects the screening of the model. The screening effectiveness depends on OOT time nodes, and the selection of the nodes needs to reflect the change of data distribution in an OOT set. However, the process often depends on expert experience, effective quantification cannot be achieved, and the screening effect is not controllable. And the OOT set is only one and can only select a model in another time division, and the change trend of data distribution in a continuous time cannot be captured.

To solve the above-mentioned drawbacks, referring to fig. 3, fig. 3 is an optional flowchart of the financial risk prediction method based on time domain information provided in the present application, and it can be understood that the steps shown in fig. 3 may be executed by various electronic devices operating a financial risk prediction apparatus based on time domain information to complete training and deployment of corresponding financial information processing models, specifically, the electronic devices may be, for example, a dedicated terminal with a financial data processing function, a server with a financial information processing model training function, or a server cluster, and implement training and deployment for financial information processing models adapted in different financial scenarios. The following is a description of the steps shown in fig. 3.

Step 301: and the financial risk prediction device based on the time domain information acquires a sample feature set matched with the financial information processing model based on the features of the target user set and the historical parameters of the financial scene.

Wherein, the sample feature set comprises sample features of different time domains. In some embodiments of the present invention, since the acquired sample features including different time domains may contain noise information, the denoising process is required, which may be implemented as follows:

and denoising the sample feature set according to the dynamic noise threshold value to form a sample feature set matched with the dynamic noise threshold value. Wherein, due to different use environments of the financial information processing model, the dynamic noise threshold value matched with the use environment of the financial information processing model is different, for example, in the financial use environment for payment and transfer through the WeChat process, the dynamic noise threshold value matched with the use environment of the financial information processing model needs to be smaller than the dynamic noise threshold value in the financial use environment for a user to purchase financial products and make financial loans through the WeChat process.

In some embodiments of the present invention, since the acquired sample features including different time domains may contain noise information, the denoising process is required, which may be implemented as follows:

and determining a fixed noise threshold corresponding to the financial information processing model, and denoising the sample feature set according to the fixed noise threshold to form a sample feature set matched with the fixed noise threshold. When the financial information processing model is solidified in a corresponding hardware mechanism, such as a financial terminal (a POS machine or a teller machine), and the use environment is a financial loan use environment for predicting risks of a target user, because the noise is relatively single, the training speed of the financial information processing model can be effectively increased and the waiting time of the user can be reduced by fixing the fixed noise threshold corresponding to the financial information processing model.

Step 302: and the financial risk prediction device based on the time domain information determines different time window lengths, processes the sample feature set according to the corresponding time window length, and determines training sample sets of different time domains matched with the financial information processing model.

Wherein the set of training samples comprises at least one set of training samples.

In some embodiments of the present invention, the sample feature set is processed based on different time window lengths, and the training sample sets of different time domains matched with the financial information processing model are determined, which may be implemented by:

determining time windows of different lengths based on the use environment of the financial information processing model; determining timestamp parameters carried by different sample data in the sample feature set; and processing the timestamp parameters carried by different sample data in the sample feature set according to a random gradient descent algorithm matched with the financial information processing model and time windows with different lengths, and determining training sample sets of different time domains matched with the financial information processing model. Optionally, the Neural Network structure in the financial information processing model may be a convolutional Neural Network (CNN convolutional Neural Network), a Deep Neural Network (DNN Deep Neural Network), a generation countermeasure Network (GAN), or the like, and the type of the Neural Network structure in the financial information processing model is not limited in the embodiment of the present invention. The neural network structure in the financial information processing model may be a neural network suitable for different target object prediction tasks, such as: a target user risk pre-judging task, a product risk evaluation task, a target user state analysis task and the like. The neural network structure in the financial information processing model can also be a neural network suitable for different application scenarios, such as: the method and the device have the advantages that the application range of the neural network structure in the financial information processing model is not limited, and the financial payment scene, the financial product purchase scene, the target user financial loan and the like are realized. Alternatively, the network structure of the neural network structure in the financial information processing model may be designed according to a computer vision task, or the network structure of the neural network structure in the financial information processing model may adopt at least a part of an existing network structure, for example: a depth residual error Network or a Visual Geometry Group Network (VGGNet Visual Geometry Group Network), etc., and the embodiment of the present invention does not limit the Network structure of the neural Network structure in the financial information processing model. Wherein, in the acquisition stage of the training sample, the acquired training data is labeled data with data acquisition time stamps and a certain time span, therefore, based on the labeled data set with a certain time span as the training sample set, by carrying out the countertraining on the samples in different time periods, training of the feature extractor network in the time domain independent financial information processing model may be accomplished, according to the random gradient descent algorithm matched with the financial information processing model and the time windows with different lengths, the timestamp parameters carried by different sample data in the sample characteristic set are processed, the method can realize mapping the different feature vector distributions of different time domains to a feature distribution space stable along with time, and complete the training of the financial information processing model through the corresponding feature space.

Step 303: and the financial risk prediction device based on the time domain information trains the financial information processing model according to the training sample sets of different time domains matched with the financial information processing model, and determines model parameters matched with the financial information processing model.

Therefore, the risk prediction of the target user in the financial scene through the financial information processing model can be realized.

In some embodiments of the present invention, training the financial information processing model according to the training sample set, and determining the model parameters adapted to the financial information processing model, may be implemented by:

training the financial information processing model according to the training sample set, and determining model parameters of a feature extractor network in the financial information processing model; training the financial information processing model according to the training sample set, and determining model parameters of a sample classifier network in the financial information processing model; and training the financial information processing model according to the training sample set, and determining model parameters of a domain classifier network in the financial information processing model. Wherein, financial information processing model training phase in this application includes: 1) feature extractor network training stage 2) training stages for the sample classifier network and the domain classifier network. Assuming that the data is divided into N time domain data sets using a preset time window,

each time domain data set has different labeled data as training samples. Of course, in other embodiments of the present invention, there is some unlabeled data that may also be added to the training samples in the data acquisition stage of the training samples, but only used as the training sample set of the domain classifier network, not used as the training sample of the sample classifier network.

In some embodiments of the present invention, training the financial information processing model according to the training sample set to determine the model parameters of the feature extractor network in the financial information processing model may be implemented by:

processing the training sample set through a feature extractor network in the financial information processing model to determine initial parameters of the feature extractor network; processing the set of training samples by the feature extractor network in response to initial parameters of the feature extractor network, determining updated parameters of the feature extractor network; and according to the updated parameters of the feature extractor network, iteratively updating the parameters of the feature extractor network through the training sample set so as to extract the feature embedded vector of each sample in the training sample set. The financial information processing model comprises a Feature Extractor network (Feature Extractor), a sample Classifier network (Classifier) and a domain Classifier network, wherein the Feature Extractor network (Feature Extractor) is responsible for extracting Feature embedding vectors required by a subsequent different classification network, and data of all different time domains share the same domain-independent Feature Extractor. The feature extractor maps each time domain sample feature into the same space. The structure is a fully-connected network, the input of the fully-connected network is the original characteristic of each sample, and the output of the fully-connected network is the characteristic embedded vector of each sample. The parameter generation network adaptively generates subsequent sample classifier network parameters and model parameters of the classifier network by taking the feature vectors as input.

In some embodiments of the present invention, the processing of the set of training samples by the feature extractor network to determine updated parameters of the feature extractor network in response to initial parameters of the feature extractor network may be implemented by: substituting different training samples in the training sample set into a loss function corresponding to the feature extractor network; and determining that the feature extractor network corresponds to the update parameter when the loss function meets the corresponding convergence condition. Wherein the content of the first and second substances,

in some embodiments of the present invention, iteratively updating the parameters of the feature extractor network through the training sample set according to the updated parameters of the feature extractor network to extract the feature embedding vector of each sample in the training sample set may be implemented by:

determining a loss function corresponding to the feature extractor network; iteratively updating the parameters of the feature extractor network according to the updated parameters of the feature extractor network; until the loss function of the feature extractor network reaches a corresponding convergence condition, and based on the parameters in the feature extractor network, a feature embedding vector of each sample in the training sample set can be extracted. The sample set processed by the feature extractor network comprises personal information features, wind control mode features, credit risk category features and matched training task labels, wherein the personal information features, the wind control mode features, the credit risk category features and the matched training task labels respectively correspond to different users in different financial business scenes. The extracted feature embedded vector of each sample can be used for a classifier network type in a financial information processing model to form a corresponding prediction result, and a domain classifier network judges time domains corresponding to different samples respectively.

In some embodiments of the invention, training the financial information processing model according to the training sample set, and determining model parameters of a sample classifier network in the financial information processing model comprises:

training samples in the financial information processing model through a sample classifier networkProcessing the set to determine initial parameters of the sample classifier network; processing the training sample set by the sample classifier network in response to initial parameters of the sample classifier network, determining updated parameters of the sample classifier network; and according to the updated parameters of the sample classifier network, iteratively updating the parameters of the sample classifier network through the training sample set so as to determine the risk prediction results of different samples based on corresponding sample labels and the feature embedded vector of each sample. Wherein different training samples may comprise a plurality of training tasks T_iA plurality of training tasks T_iProvided with respective different class sample classifier networks P_iDifferent class sample classifier networks P_iEach training task T corresponding to different financial business scenes_iA plurality of data samples are included, each data sample including a personal information feature of a respective user, a wind control pattern feature, a credit risk category of the respective user, and a task label indicating a training task to which the data sample belongs. Illustratively, the personal information characteristic and the wind control pattern characteristic of the user may include personal basic information (such as sex, age, occupation, income, standing address, tax payment information, credit investigation information and the like of the user), social behavior information (parameters such as login frequency, last login time, registration time, activity degree and the like of social software such as WeChat), financial behavior information (such as consumption payment information, historical loan information, credit investigation information, bank grade information and the like), used equipment information and the like. The credit risk categories for a user may simply include, for example, high risk users and medium risk users and low risk users, or be represented in more detail in terms of credit scores, credit ratings, and the like. The task label may be payment (such as defining the task label according to payment time, payment object type or payment channel information, payment amount, etc.), loan (such as defining the task label according to loan type, loan amount, payment time limit, payment interest, etc.), financing (such as defining any one of financing product type, financing supervisor information, risk level information, fund closure time limit, etc.)Service tags), task tags in different financial business scenarios using environment and noise information (such as a deployment terminal environment of a financial information processing model, financial business environment, fixed noise, dynamic noise).

In some embodiments of the present invention, the processing of the training sample set by the sample classifier network to determine the updated parameters of the sample classifier network in response to the initial parameters of the sample classifier network may be implemented by:

substituting different training samples in the training sample set into a loss function corresponding to the sample classifier network; and determining that the sample classifier network corresponds to the update parameters when the loss function meets the corresponding convergence condition.

In some embodiments of the present invention, according to the updated parameters of the sample classifier network, iteratively updating the parameters of the sample classifier network through the training sample set to determine the risk prediction results of different samples based on the corresponding sample labels and the feature embedding vectors of each sample, may be implemented by:

determining a loss function corresponding to the sample classifier network; iteratively updating the parameters of the sample classifier network according to the updated parameters of the sample classifier network; until the loss function of the sample classifier network reaches the corresponding convergence condition, and based on the corresponding sample label and the feature embedding vector of each sample, determining the risk prediction results of different samples. The sample classifier network in the financial information processing model can perform identification, classification and prediction on the marked samples in all domains, and optionally, the structure of the sample classifier network is a fully-connected network. And embedding the characteristic of each marked sample into a vector, and outputting the prediction category of the sample to prompt whether the current user is at risk of financial loan.

In some embodiments of the present invention, training the financial information processing model according to the training sample set to determine the model parameters of the domain classifier network in the financial information processing model may be implemented by:

in response to initial parameters of the domain classifier network, processing the training sample set through the domain classifier network to determine updated parameters of the domain classifier network; and according to the updated parameters of the domain classifier network, performing iterative updating on the parameters of the domain classifier network through the training sample set so as to judge the time domains corresponding to different samples in the sample set respectively. The domain identification classifier network can perform domain discrimination prediction on all samples collected in the continuous set, and judge which time domain the current sample comes from. The structure of the domain identification classifier network is a full-connection network, the input of the domain identification classifier network is a feature embedding vector of each marked sample, and the output is a prediction result of domain discrimination. And fitting the real domain relation of the domain by constraining the w-distance, namely if the two classes come from the same time domain, further, the features of different time domains can be mapped to the embedded vector on the same table task-space in a supervision manner.

In some embodiments of the present invention, in response to the initial parameter of the domain classifier network, the training sample set is processed by the domain classifier network, and the updated parameter of the domain classifier network is determined, which may be implemented by:

substituting different training samples in the training sample set into a loss function corresponding to the domain classifier network; and determining that the domain classifier network corresponds to the update parameters when the loss function meets the corresponding convergence condition.

In some embodiments of the present invention, iteratively updating the parameters of the domain classifier network through the training sample set according to the updated parameters of the domain classifier network to extract the feature embedding vector of each sample in the training sample set may be implemented by:

determining a loss function corresponding to the domain classifier network; iteratively updating the parameters of the domain classifier network according to the updated parameters of the domain classifier network; until the loss function of the domain classifier network reaches the corresponding convergence condition, and based on the parameters in the domain classifier network, the time domains corresponding to different samples in the sample set can be judged. In order to overcome the defects of the conventional financial information processing method described in the preamble embodiment, the feature extractor network and the domain classifier network may be trained in a countermeasure mode. The information extracted by the feature extractor network in forward propagation is transmitted into a domain classifier network, and the domain classifier network judges which time domain the transmitted information belongs to and calculates the domain classification loss. The goal of a domain classifier network is to distinguish as much as possible which time domain the features of the input come in. In the back propagation process, a gradient reversal layer (gradient reverse layer) between the domain classifier network and the feature extractor network makes the training target of the feature extractor network opposite to that of the domain classifier network, that is, the features that the feature extractor network wants to output make the domain classifier network unable to correctly judge which domain the information comes from. The confrontation relationship finally enables the domain classifier network not to correctly distinguish the received information, and the feature extractor network successfully mixes different time domain samples in a certain public feature space, so that the bloom and robustness of the trained financial information processing model are improved, and the reduction of the prediction capability caused by the long-time operation of the financial information processing is avoided.

In some embodiments of the present invention, time stamp parameters carried by different sample data in a sample feature set may also be processed according to a random gradient descent algorithm matched with the financial information processing model and time windows of different lengths, so as to determine a test sample set of different time domains matched with the financial information processing model; and processing different test samples in the test sample set through the financial information processing model so as to test the risk prediction result output by the financial information processing model through the test sample set. Referring to fig. 4, fig. 4 is a schematic diagram of an alternative model training scheme in an embodiment of the present invention, wherein the financial information processing model may be trained by an error back propagation algorithm. The risk prediction for the target user is the need to use historical data as a training set and a validation set. The training set validation set is constructed as shown in fig. 4. According to different timestamp parameters, the training set (2 time domains) is used earlier, the optimal model is selected by using the training set as the verification set in the latest period, and finally the future is predicted based on the existing data.

Further, the trained financial information processing model can be deployed in different servers, server groups and financial blockchain networks.

With continuing reference to fig. 5, fig. 5 is an alternative flow chart of the financial risk prediction method based on time domain information provided in the present application, and it can be understood that the steps shown in fig. 5 can be executed by various electronic devices operating a financial risk prediction apparatus based on time domain information, such as a dedicated terminal with a financial data processing function, a server with a financial information processing function, or a server cluster, to implement financial information processing using financial information processing models adapted in different deployed financial scenarios. The following is a description of the steps shown in fig. 5.

Step 501: and the financial risk prediction device based on the time domain information receives a financial risk judgment request sent by the client.

Step 502: the financial risk prediction device based on the time domain information responds to a financial risk prediction request, and obtains unlabeled financial information sample characteristics matched with the target user.

Step 503: and the financial risk prediction device based on the time domain information determines different financial service scenes respectively corresponding to the unlabeled financial information sample characteristics matched with the target user.

Step 504: the financial risk prediction device based on time domain information determines dynamic noise matched with different financial business scenes based on the different financial business scenes; and based on the dynamic noise, carrying out dynamic denoising processing on the unlabeled financial information sample characteristics matched with the target user to form an unlabeled sample characteristic set matched with the corresponding financial business scene.

Step 505: and the financial risk prediction device based on the time domain information processes the non-labeled financial information sample characteristics through a financial information processing model to form a financial risk prediction result of the target user.

Step 506: and the financial risk prediction device based on the time domain information outputs the financial risk prediction result of the target user.

Step 507: and determining to execute corresponding financial debit and credit amounts based on the risk prediction results of different target users.

In some embodiments of the present invention, different financial service scenarios respectively corresponding to the unlabeled financial information sample features matching the target user are determined;

determining dynamic noise matched with the different financial business scenes based on the different financial business scenes; and based on the dynamic noise, dynamically denoising the unlabeled financial information sample characteristics matched with the target users to form a sample characteristic set matched with the corresponding financial business scene, wherein each sample comprises the personal characteristics of the corresponding target user and the credit risk category characteristics of the target user. When a user who wants to transact related financial business by a target user of a financial APP uses client equipment to access services provided by a client server of an enterprise, the client server can submit personal characteristic information of the user to a trained financial information processing model to obtain a corresponding prediction result for risk prediction, so that the credit risk category of the user is obtained. The financial information processing model can assist the financial platform or the lender to judge whether to provide financial business service for the user or assist different lenders in the financial platform to perform different management on users with different credit risk types.

Meanwhile, the scheme of the application can be realized through a financial APP in practical application, and meanwhile, the scheme of the application can also be realized through a WeChat applet, so that a user can quickly predict financial information of different target objects through a financial information processing model deployed in a block chain network when changing a terminal through a financial block chain.

Specifically, the target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier may be sent to a blockchain network, so that a node of the blockchain network fills the target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier into a new block, and when the new block consensus is consistent, the new block is added to the tail of the blockchain.

The embodiment of the present invention may be implemented by combining a Cloud technology, where the Cloud technology (Cloud technology) is a hosting technology for unifying series resources such as hardware, software, and a network in a wide area network or a local area network to implement calculation, storage, processing, and sharing of data, and may also be understood as a generic term of a network technology, an information technology, an integration technology, a management platform technology, an application technology, and the like applied based on a Cloud computing business model. Background services of the technical network system require a large amount of computing and storage resources, such as video websites, photo-like websites and more portal websites, so cloud technology needs to be supported by cloud computing.

It should be noted that cloud computing is a computing mode, and distributes computing tasks on a resource pool formed by a large number of computers, so that various application systems can obtain computing power, storage space and information services as required. The network that provides the resources is referred to as the "cloud". Resources in the "cloud" appear to the user as being infinitely expandable and available at any time, available on demand, expandable at any time, and paid for on-demand. As a basic capability provider of cloud computing, a cloud computing resource pool platform, which is called an Infrastructure as a Service (IaaS) for short, is established, and multiple types of virtual resources are deployed in a resource pool and are used by external clients selectively. The cloud computing resource pool mainly comprises: a computing device (which may be a virtualized machine, including an operating system), a storage device, and a network device.

As shown in fig. 1, the target object determining method provided in the embodiment of the present invention may be implemented by corresponding cloud devices, for example: the terminals (including the terminal 10-1 and the terminal 10-2) are connected to the server 200 located at the cloud end through a network 300, and the network 300 may be a wide area network or a local area network, or a combination of the two. It should be noted that the server 200 may be a physical device or a virtualized device.

In some embodiments of the invention the method further comprises:

In some embodiments of the present invention, the query request may be further analyzed to obtain a corresponding object identifier in response to the query request; acquiring authority information in a target block in a block chain network according to the object identifier; checking the matching of the authority information and the object identification; when the authority information is matched with the object identification, acquiring a corresponding target user identification, a model parameter of a financial information processing model and a financial scene identification in the block chain network; and responding to the query instruction, pushing the acquired corresponding financial information and the target object matched with the basic object to a corresponding client so as to enable the client to acquire the corresponding financial information stored in the blockchain network and the target object matched with the basic object.

Referring to fig. 6, fig. 6 is a schematic structural diagram of a financial risk prediction apparatus 100 based on time domain information according to an embodiment of the present invention, which includes a blockchain network 200 (illustratively, a consensus node 210-1, a consensus node 210-2, and a consensus node 210-3), an authentication center 300, a business entity 400, and a business entity 500, which are respectively described below.

The type of blockchain network 200 is flexible and may be, for example, any of a public chain, a private chain, or a federation chain. Taking a public link as an example, electronic devices such as user terminals and servers of any service entity can access the blockchain network 200 without authorization; taking a federation chain as an example, an electronic device (e.g., a terminal/server) under the jurisdiction of a service entity after obtaining authorization may access the blockchain network 200, and at this time, become a client node in the blockchain network 200.

In some embodiments, the client node may act as a mere watcher of the blockchain network 200, i.e., provides functionality to support a business entity to initiate a transaction (e.g., for uplink storage of data or querying of data on a chain), and may be implemented by default or selectively (e.g., depending on the specific business requirements of the business entity) with respect to the functions of the consensus node 210 of the blockchain network 200, such as a ranking function, a consensus service, and an accounting function, etc. Therefore, the data and the service processing logic of the service subject can be migrated into the block chain network 200 to the maximum extent, and the credibility and traceability of the data and service processing process are realized through the block chain network 200.

The consensus nodes in blockchain network 200 receive transactions submitted from client nodes (e.g., client node 410 shown in fig. 6 as belonging to business entity 400 and client node 510 shown in fig. 6 as belonging to server 500) of different business entities (e.g., business entity 400 and business entity 500 shown in fig. 6), perform the transactions to update the ledger or query the ledger, and various intermediate or final results of performing the transactions may be returned for display in the business entity's client nodes.

For example, the client node 410/510 may subscribe to events of interest in the blockchain network 200, such as transactions occurring in a particular organization/channel in the blockchain network 200, and the corresponding transaction notifications are pushed by the consensus node 210 to the client node 410/510, thereby triggering the corresponding business logic in the client node 410/510.

An exemplary application of the blockchain network is described below, taking an example that a plurality of business entities access the blockchain network to realize management and processing of financial information.

Referring to fig. 6, a plurality of business entities involved in the management process, such as business entity 400, may be an artificial intelligence-based financial risk prediction apparatus based on time domain information, and business entity 500 may be a display system with a financial information display (operation) function, and registers from the certificate authority 300 to obtain respective digital certificates, including the public key of the business entity and the digital signature signed by the certificate authority 300 for the public key and identity information of the business entity, for attaching to the transaction together with the digital signature of the business entity for the transaction, and sending to the blockchain network, for the blockchain network to take out the digital certificate and signature from the transaction, verify the authenticity of the message (i.e. whether it has not been tampered with) and the identity information of the service entity sending the message, and the blockchain network will verify according to the identity, for example whether it has the right to initiate the transaction. Clients running on electronic devices (e.g., terminals or servers) hosted by the business entity may request access from the blockchain network 200 to become client nodes.

The client node 410 of the business entity 400 is configured to obtain financial information matching the target user in response to the financial information prediction request; determining a knowledge-graph of a target object in the financial information prediction request based on the financial information; determining a trend hidden variable matched with the target object based on the financial information; determining a graph neural network in a financial information processing model according to the knowledge graph of the target object and the trend hidden variable matched with the target object; and determining the change trend of the corresponding target object based on the graph neural network in the financial information processing model so as to realize the response of the financial information prediction request through the change trend of the target object. And sending the target user identification, the model parameters of the financial information processing model and the financial scene identification to the block chain network 200.

The target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier may be sent to the blockchain network 200, a service logic may be set in the client node 410 in advance, and when corresponding financial information is formed, the client node 410 automatically sends the target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier to the blockchain network 200, or a service person of the service agent 400 logs in the client node 410, manually packages the target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier, and sends the same to the blockchain network 200. During sending, the client node 410 generates a transaction corresponding to the update operation according to the target user identifier, the model parameters of the financial information processing model, and the financial scenario identifier, specifies an intelligent contract that needs to be invoked to implement the update operation, and parameters transferred to the intelligent contract in the transaction, and also carries a digital certificate of the client node 410 and a signed digital signature (for example, a digest of the transaction is encrypted using a private key in the digital certificate of the client node 410), and broadcasts the transaction to the consensus node 210 in the blockchain network 200.

When the transaction is received in the consensus node 210 in the blockchain network 200, the digital certificate and the digital signature carried by the transaction are verified, and after the verification is successful, whether the service agent 400 has the transaction right or not is determined according to the identity of the service agent 400 carried in the transaction, and the transaction fails due to any verification judgment of the digital signature and the right verification. After successful verification, the consensus node 210 signs its own digital signature (e.g., by encrypting a digest of the transaction using the private key of the consensus node 210-1) and continues to broadcast in the blockchain network 200.

After receiving the transaction successfully verified, the consensus node 210 in the blockchain network 200 fills the transaction into a new block and broadcasts the new block. When a new block is broadcasted by the consensus node 210 in the block chain network 200, performing a consensus process on the new block, if the consensus is successful, adding the new block to the tail of the block chain stored in the new block, updating the state database according to a transaction result, and executing a transaction in the new block: and for the transaction of submitting and updating the target user identification, the model parameters of the financial information processing model and the financial scene identification, adding the key value pair comprising the target user identification, the model parameters of the financial information processing model and the financial scene identification into the state database.

A service person of the service agent 500 logs in the client node 510, inputs a target user identifier, a model parameter of a financial information processing model, and a financial scenario identifier query request, the client node 510 generates a transaction corresponding to an update operation/query operation according to the target user identifier, the model parameter of the financial information processing model, and the financial scenario identifier query request, specifies an intelligent contract that needs to be invoked to implement the update operation/query operation and parameters transferred to the intelligent contract in the transaction, and the transaction also carries a digital certificate of the client node 510 and a signed digital signature (for example, a digest of the transaction is encrypted by using a private key in the digital certificate of the client node 510), and broadcasts the transaction to the common identification node 210 in the blockchain network 200.

After receiving the transaction in the consensus node 210 in the blockchain network 200, verifying the transaction, filling the block and making the consensus consistent, adding the filled new block to the tail of the blockchain stored in the new block, updating the state database according to the transaction result, and executing the transaction in the new block: updating the key value pair corresponding to the target user in the state database according to different target user identifications for the submitted transaction for updating the target user identification of a certain text, the model parameters of the financial information processing model and the financial scene identification; and for the submitted transaction for inquiring a certain target user, inquiring the key value pair corresponding to the target user from the state database, and returning a transaction result.

It is noted that fig. 6 illustrates a process of directly linking the target user identifier, the model parameters of the financial information processing model, and the financial scenario identifier, but in other embodiments, for a large data volume of the target user, the client node 410 may link the hash of the target user and the hash of the corresponding financial information in pairs, and store the original target user and the corresponding target financial information in a distributed file system or a database. After obtaining the target user and the corresponding target financial information from the distributed file system or the database, the client node 510 may perform a check in combination with the corresponding hash in the blockchain network 200, thereby reducing the workload of the uplink operation.

As an example of a block chain, referring to fig. 7, fig. 7 is a schematic structural diagram of a block chain in a block chain network 200 according to an embodiment of the present invention, where a header of each block may include hash values of all transactions in the block and also include hash values of all transactions in a previous block, a record of a newly generated transaction is filled in the block and is added to a tail of the block chain after being identified by nodes in the block chain network, so as to form a chain growth, and a chain structure based on hash values between blocks ensures tamper resistance and forgery prevention of transactions in the block. The target users stored in the blockchain network can be financial information in different financial scenes, and the target users can be shared among different nodes by storing the target users in the blockchain network.

An exemplary functional architecture of a block chain network provided in the embodiment of the present invention is described below, referring to fig. 8, fig. 8 is a functional architecture schematic diagram of a block chain network 200 provided in the embodiment of the present invention, which includes an application layer 201, a consensus layer 202, a network layer 203, a data layer 204, and a resource layer 205, which are described below respectively.

The resource layer 205 encapsulates the computing, storage, and communication resources that implement each of the consensus nodes 210 in the blockchain network 200.

The data layer 204 encapsulates various data structures that implement the ledger, including blockchains implemented in files in a file system, state databases of the key-value type, and presence certificates (e.g., hash trees of transactions in blocks).

The network layer 203 encapsulates the functions of a Point-to-Point (P2P) network protocol, a data propagation mechanism and a data verification mechanism, an access authentication mechanism and service agent identity management.

Wherein, the P2P network protocol implements communication between the consensus nodes 210 in the blockchain network 200, the data propagation mechanism ensures propagation of transactions in the blockchain network 200, and the data verification mechanism is used for implementing reliability of data transmission between the consensus nodes 210 based on cryptography methods (e.g., digital certificates, digital signatures, public/private key pairs); the access authentication mechanism is used for authenticating the identity of the service subject added into the block chain network 200 according to an actual service scene, and endowing the service subject with the authority of accessing the block chain network 200 when the authentication is passed; the business entity identity management is used to store the identity of the business entity that is allowed to access blockchain network 200, as well as the permissions (e.g., the types of transactions that can be initiated).

The consensus layer 202 encapsulates the functions of the mechanism for the nodes 210 in the blockchain network 200 to agree on a block (i.e., a consensus mechanism), transaction management, and ledger management. The consensus mechanism comprises consensus algorithms such as POS, POW and DPOS, and the pluggable consensus algorithm is supported.

The transaction management is configured to verify a digital signature carried in the transaction received by the consensus node 210, verify identity information of the service entity, and determine whether the service entity has the right to perform the transaction according to the identity information (read related information from the identity management of the service entity); for the service agents authorized to access the blockchain network 200, the service agents all have digital certificates issued by the certificate authority, and the service agents sign the submitted transactions by using private keys in the digital certificates of the service agents, so that the legal identities of the service agents are declared.

The ledger administration is used to maintain blockchains and state databases. For the block with the consensus, adding the block to the tail of the block chain; executing the transaction in the acquired consensus block, updating the key-value pairs in the state database when the transaction comprises an update operation, querying the key-value pairs in the state database when the transaction comprises a query operation and returning a query result to the client node of the business entity. Supporting query operations for multiple dimensions of a state database, comprising: querying the block based on the block vector number (e.g., hash value of the transaction); inquiring the block according to the block hash value; inquiring a block according to the transaction vector number; inquiring the transaction according to the transaction vector number; inquiring account data of a business main body according to an account (vector number) of the business main body; and inquiring the block chain in the channel according to the channel name.

The application layer 201 encapsulates various services that the blockchain network can implement, including tracing, crediting, and verifying transactions.

The following objective is stocks, and the financial scenario is stock trading, for example, a description is given to the financial risk prediction method based on time domain information provided by the present invention, where the types of financial resource configuration processes such as funds, stocks, etc. acquired by a user from a corresponding server 200 through a network 300 by using terminals (including a terminal 10-1 and a terminal 10-2) shown in fig. 1 may be the same or different, for example, platforms such as change, stock trading APP, etc., a user may purchase stocks conveniently and quickly to obtain profits, so that the prediction information of stocks is obtained in time, changes of a stock market index are determined, and it is helpful to improve the user experience.

In the following, the prediction of the target user who needs to make a financial loan in the financial loan usage scenario is taken as an example, the time domain information-based financial risk prediction method and the time domain information-based financial risk prediction method provided by the present application will be described, referring to fig. 9, fig. 9 is a schematic diagram of a front end display of a financial risk prediction method based on time domain information provided by the present application, wherein the terminals (such as terminal 10-1 and terminal 10-2 in figure 1) are provided with clients capable of displaying software for making financial debits accordingly, for example, a client or a plug-in for virtual resources or physical resources to perform financial activities or virtual resource lending, and a user may obtain a loan to a financial institution or a platform through a corresponding client (for example, a process of making a financial payment through WeChat payment or making a fund loan to purchase an article in WeChat); the terminal is connected to the server 200 through a network 300, and the network 300 may be a wide area network or a local area network, or a combination of the two, and uses a wireless link to realize data transmission. Servers (e.g., the server in fig. 1) of enterprises that provide financial transactions such as payment, loan, financing, etc., such as banks, securities, mutual funds, P2P, etc. When a user who needs to transact related financial business uses client equipment to access services provided by a client server of an enterprise, the client server can submit personal information characteristics, wind control mode characteristics and information of the user to a trained financial information processing model to obtain a corresponding prediction result for risk prediction so as to obtain a credit risk category of the user. The financial information processing model can assist the financial platform or the lender to judge whether to provide financial business service for the user or assist different lenders in the financial platform to perform different management on users with different credit risk types.

Referring to fig. 10, fig. 10 is a schematic diagram illustrating a usage process of the financial risk prediction method based on time domain information provided by the present application, wherein the financial risk prediction method based on time domain information provided by the present application includes the following steps:

step 1001: the server acquires the characteristics of the financial sample with the label;

step 1002: and the server divides the financial samples to form a training sample set and a testing sample set.

The time window length of sample division can be set for the financial sample characteristics, and all financial sample characteristics are divided into training sample sets of at least 2 time domains based on data acquisition time.

Step 1003: the server trains the financial information processing model through the labeled financial sample training set, and evaluates the training effect of the model in the test set.

Therein, a financial information processing model training process based on the present application is described. The input data for model training includes all samples from N time domains, as shown in steps 1) -9) of the algorithm, with each iteration of pre-training from all samples from N time domains

Randomly sampling minipatch samples of two time domains, and performing feature extractor network by adopting SGD gradient descent method

And sample classifier network

(an alternative form of a sample classifier network) the parameter update is performed. The training device of the financial information processing model obtains N time domainsOf the data set

Then, the executing step of the executed training algorithm of the financial information processing model comprises the following steps:

1) initializing DNN network parameters

；

2) Repeating;

3) sampling two time domains

And

respectively from

And

middle sampling

，

，

4) Computing sample feature extractor

In that

The feature embedding vector of each sample, wherein the embedding vector represents:

wherein

；

5) Computing financial information processing model classification network

Prediction of results on minibratchX

Wherein, in the step (A),

；

6) calculating the predicted result of labeled samples on minibochcx

Loss of (2):

wherein the content of the first and second substances,

；

7) computing domain identification network

Loss on minibtchxx, where,

，

；

8) determining network parameters of a gradient update feature extractor and a sample classifier:

wherein the feature extractor updates:

updating a sample classifier:

(ii) a Domain classificationUpdating the device:

；

9) updating is carried out until

And

and

and (6) converging.

The following describes the algorithms shown in the preceding steps, respectively, 1) feature extractor network training (algorithm step 4)): feature extractor (Feature extractor)

And the device is responsible for extracting the feature embedding vectors required by the subsequent classification network, and the data of all different time domains share the same domain-independent feature extractor. The feature extractor maps each time domain sample feature into the same space.

The structure of (1) is a fully-connected network, the input of which is the original characteristic of each sample, and the output is the characteristic embedding vector of each sample.

2) Sample classifier network (algorithm step 5) and step 6)): fraud detection is a sample classification algorithm, a sample classifier network

Is responsible for fraud identification classification prediction of marked samples in all domains,

the structure of (1) is a fully connected network. Its input is the feature embedding vector of each labeled sample, and its output is the prediction category of the sample.

3) Domain identification classification network (algorithm step 7)): domain identification classification network

Is responsible for carrying out domain discrimination prediction on all samples and judging that the current sample comes from a time domain T_iOr T_j. The domain identification classification network is structured as a fully connected network. The input of the method is a characteristic embedding vector of each marked sample, and the output is a prediction result of domain discrimination. Fitting a true domain relationship of a domain by constraining w-distance distances

I.e. two classes from the same time domain

From different time domains

Features of different time domains can be mapped to embedded vectors on the same table task-space supervised.

Step 1004: the server deploys the trained financial information processing model.

Step 1005: and the server responds to the processing request, loads a financial information processing model to process the non-tag financial sample in the financial scene, and determines a risk prediction result of the target user.

Referring to fig. 11, fig. 11 is a schematic front-end display diagram of a financial risk prediction method based on time domain information according to the present application, in which a terminal (e.g., the terminal 10-1 and the terminal 10-2 in fig. 1) is provided with a client capable of displaying software for performing financial loan, and after a target user issues a financial loan request through the client of the software for performing financial loan provided by the terminal, different financial information processing models deployed in different servers may respectively evaluate risks of the target user to determine a risk level of the target user and determine whether to loan the target user.

Therefore, the defect that all marked samples and unmarked samples cannot be fully utilized caused by the fact that a traditional machine learning screening model is screened through a marked OOT data set isolated in time is overcome. The method has the advantages that all marked samples can be fully utilized for training, the real-time label-free samples can be fully utilized for carrying out iterative updating on the model, the time for marking the samples and the marking cost are reduced, and the model training efficiency is improved. Meanwhile, the influence of selecting a single OOT time node on the financial information processing model is overcome, and the data sets are divided by fully utilizing a plurality of time nodes, so that the financial information processing model can fully learn the continuous variation trend of data in the time dimension. Meanwhile, the influence of a single OOT time node on the model can be reduced, the dependence on unquantized modeling standards such as expert experience can be reduced, and the training standard and the training process of the financial information processing model of the financial wind control can be effectively normalized and quantized. Furthermore, the financial information processing model can accurately identify the high-risk fraudulent user by using a time domain countermeasure method, the effect can be stable along with the change of data distribution in the time dimension, and the life cycle of the financial information processing model is effectively prolonged. In a financial wind control use scene, continuous, efficient and stable operation of the business can be effectively guaranteed, and the problem that risks brought to the business by frequent updating of the model are unstable is avoided.

The beneficial technical effects are as follows:

The above description is only exemplary of the present invention and should not be taken as limiting the scope of the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. A financial risk prediction method based on time domain information, the method comprising:

and outputting the financial risk prediction result of the target user.

2. The method of claim 1, further comprising:

determining different time window lengths, processing the sample feature set according to the corresponding time window lengths, and determining training sample sets of different time domains matched with the financial information processing model, wherein the training sample sets comprise at least one group of training samples;

3. The method of claim 2, wherein determining different time window lengths and processing the sample feature sets according to the corresponding time window lengths, and determining training sample sets of different time domains matching the financial information processing model comprises:

determining the use environment of the financial information processing model, and determining time windows with different lengths according to the use environment of the financial information processing model;

determining timestamp parameters carried by different sample data in the sample feature set;

and processing the timestamp parameters carried by different sample data in the sample feature set according to a random gradient descent algorithm matched with the financial information processing model and time windows with different lengths, and determining training sample sets of different time domains matched with the financial information processing model.

4. The method of claim 2, wherein training the financial information processing model based on the set of training samples to determine model parameters that fit the financial information processing model comprises:

training the financial information processing model according to the training sample set, and determining model parameters of a feature extractor network in the financial information processing model;

training the financial information processing model according to the training sample set, and determining model parameters of a sample classifier network in the financial information processing model;

and training the financial information processing model according to the training sample set, and determining model parameters of a domain classifier network in the financial information processing model.

5. The method of claim 4, wherein training the financial information processing model based on the set of training samples to determine model parameters of a feature extractor network in the financial information processing model comprises:

processing the training sample set through a feature extractor network in the financial information processing model to determine initial parameters of the feature extractor network;

processing the set of training samples by the feature extractor network in response to initial parameters of the feature extractor network, determining updated parameters of the feature extractor network;

and according to the updated parameters of the feature extractor network, iteratively updating the parameters of the feature extractor network through the training sample set so as to extract the feature embedded vector of each sample in the training sample set.

6. The method of claim 5, wherein iteratively updating the parameters of the feature extractor network through the set of training samples according to the updated parameters of the feature extractor network to extract the feature embedding vector for each sample in the set of training samples comprises:

determining a loss function corresponding to the feature extractor network;

iteratively updating the parameters of the feature extractor network according to the updated parameters of the feature extractor network; until the loss function of the feature extractor network reaches a corresponding convergence condition, and based on the parameters in the feature extractor network, a feature embedding vector of each sample in the training sample set can be extracted.

7. The method of claim 5, wherein training the financial information handling model based on the set of training samples to determine model parameters of a network of sample classifiers in the financial information handling model comprises:

processing the training sample set through a sample classifier network in the financial information processing model to determine initial parameters of the sample classifier network;

processing the training sample set by the sample classifier network in response to initial parameters of the sample classifier network, determining updated parameters of the sample classifier network;

and according to the updated parameters of the sample classifier network, iteratively updating the parameters of the sample classifier network through the training sample set so as to determine the risk prediction results of different samples based on corresponding sample labels and the feature embedded vector of each sample.

8. The method of claim 7, wherein the processing the set of training samples by the sample classifier network in response to initial parameters of the sample classifier network to determine updated parameters of the sample classifier network comprises:

substituting different training samples in the training sample set into a loss function corresponding to the sample classifier network;

and determining that the sample classifier network corresponds to the update parameters when the loss function meets the corresponding convergence condition.

9. The method of claim 7, wherein iteratively updating the parameters of the sample classifier network through the training sample set according to the updated parameters of the sample classifier network to determine risk prediction results of different samples based on corresponding sample labels and feature embedding vectors of each sample comprises:

determining a loss function corresponding to the sample classifier network;

iteratively updating the parameters of the sample classifier network according to the updated parameters of the sample classifier network; until the loss function of the sample classifier network reaches the corresponding convergence condition, and based on the corresponding sample label and the feature embedding vector of each sample, determining the risk prediction results of different samples.

10. The method of claim 3, wherein training the financial information handling model based on the set of training samples to determine model parameters for a domain classifier network in the financial information handling model comprises:

in response to initial parameters of the domain classifier network, processing the training sample set through the domain classifier network to determine updated parameters of the domain classifier network;

and according to the updated parameters of the domain classifier network, performing iterative updating on the parameters of the domain classifier network through the training sample set so as to judge the time domains corresponding to different samples in the sample set respectively.

11. The method of claim 10, further comprising:

substituting different training samples in the training sample set into a loss function corresponding to the domain classifier network;

determining that the domain classifier network corresponds to the update parameter when the loss function satisfies a corresponding convergence condition;

determining a loss function corresponding to the domain classifier network;

iteratively updating the parameters of the domain classifier network according to the updated parameters of the domain classifier network; until the loss function of the domain classifier network reaches the corresponding convergence condition, and based on the parameters in the domain classifier network, the time domains corresponding to different samples in the sample set can be judged.

12. The method of claim 2, further comprising:

processing timestamp parameters carried by different sample data in a sample characteristic set according to a random gradient descent algorithm matched with the financial information processing model and time windows with different lengths, and determining test sample sets of different time domains matched with the financial information processing model;

and processing different test samples in the test sample set through the financial information processing model so as to test the risk prediction result output by the financial information processing model through the test sample set.

13. An apparatus for predicting financial risk based on time domain information, the apparatus comprising:

the information processing module is used for determining dynamic noise matched with the different financial business scenes based on the different financial business scenes;

the information processing module is used for carrying out dynamic denoising processing on the unlabeled financial information sample characteristics matched with the target user based on the dynamic noise so as to form an unlabeled sample characteristic set matched with a corresponding financial business scene;

14. An electronic device, characterized in that the electronic device comprises:

a memory for storing executable instructions;

a processor configured to execute the executable instructions stored in the memory to implement the method of time domain information based financial risk prediction of any one of claims 1 to 12.

15. A computer-readable storage medium storing executable instructions, wherein the executable instructions when executed by a processor implement the method for financial risk prediction based on time domain information of any one of claims 1 to 12.