CN111210347A

CN111210347A - Transaction risk early warning method, device, equipment and storage medium

Info

Publication number: CN111210347A
Application number: CN201911336540.6A
Authority: CN
Inventors: 骆宗伟; 何睿智; 李文龙
Original assignee: Southwest University of Science and Technology
Current assignee: Southwest University of Science and Technology; Southern University of Science and Technology
Priority date: 2019-12-23
Filing date: 2019-12-23
Publication date: 2020-05-29

Abstract

The application is applicable to the technical field of computers, and provides a transaction risk early warning method, a transaction risk early warning device, transaction risk early warning equipment and a storage medium. The method comprises the steps of acquiring time sequence data of various transaction signals; inputting the time sequence data into a transaction strategy feature prediction model to obtain transaction strategy applicable features corresponding to each time window; and inputting the applicable characteristics of the transaction strategy of each time window into a trained Bayesian classifier to obtain a transaction strategy reference line and an early warning index value corresponding to each time window. According to the transaction risk early warning method, the transaction strategy feature prediction model is obtained based on long-time memory neural network training, the time sequence data comprises the change features of each transaction signal on a plurality of continuous time windows, the association relation between the transaction signals of the financial market and the applicable features of the transaction strategy in different time periods has predictability, and the reliability of transaction strategy risk early warning is improved.

Description

Transaction risk early warning method, device, equipment and storage medium

Technical Field

The application belongs to the technical field of computers, and particularly relates to a transaction risk early warning method, a transaction risk early warning device, transaction risk early warning equipment and a storage medium.

Background

Financial transaction risk early warning plays an important role in the optimal configuration of enterprise resources. The early warning of the financial transaction risk can not only bring great economic loss to enterprises, but also seriously affect the international market competitiveness of the enterprises, and more seriously affect the integral image and economic strength of the country.

The financial transaction risk early warning means that under the current transaction signal, early warning is provided for risks which may appear after the transaction strategy is adopted and implemented by a decision maker. Different transaction strategies have their application; at present, financial transaction risk early warning is mainly based on a specific transaction strategy adopted in a static financial transaction market, and the applicability problem of the transaction strategy is not considered; because financial market changes are uncommon, the same set of early warning rules cannot deal with different financial market conditions, and the risk early warning reliability is low.

Disclosure of Invention

In view of this, embodiments of the present application provide a transaction risk early warning method, apparatus, device and storage medium, so as to solve the technical problem in the prior art that the transaction risk early warning reliability is low.

In a first aspect, an embodiment of the present application provides a transaction risk early warning method, including:

acquiring time sequence data of various transaction signals; the time series data comprises a variation characteristic of each transaction signal over a plurality of successive time windows;

inputting the time sequence data into a transaction strategy feature prediction model to obtain transaction strategy applicable features corresponding to each time window; the transaction strategy feature prediction model is obtained based on long-time and short-time memory neural network training;

and inputting the applicable characteristics of the transaction strategy of each time window into a trained Bayesian classifier to obtain a transaction strategy reference line and an early warning index value corresponding to each time window.

In a second aspect, an embodiment of the present application provides a transaction risk early warning apparatus, including:

the acquisition module is used for acquiring time sequence data of various transaction signals; the time series data comprises a variation characteristic of each transaction signal over a plurality of successive time windows;

the prediction module is used for inputting the time series data into a transaction strategy characteristic prediction model to obtain transaction strategy applicable characteristics corresponding to each time window; the transaction strategy feature prediction model is obtained based on long-time and short-time memory neural network training;

and the early warning module is used for inputting the applicable characteristics of the transaction strategy of each time window into the trained Bayesian classifier to obtain a transaction strategy reference line and an early warning index value corresponding to each time window.

In a third aspect, an embodiment of the present application provides a transaction risk early warning device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of any one of the methods in the first aspect when executing the computer program.

In a fourth aspect, the present application provides a computer-readable storage medium, where a computer program is stored, and when executed by a processor, the computer program implements the steps of any one of the methods in the first aspect.

In a fifth aspect, the present application provides a computer program product, which when run on a terminal device, causes the terminal device to execute the method of any one of the above first aspects.

The transaction risk early warning method provided by the embodiment of the application obtains the change characteristics of various transaction signals in the financial market on a plurality of continuous time windows; and inputting the characteristics of various transaction signals which change on a plurality of continuous time windows into a transaction strategy characteristic prediction model to obtain transaction strategy applicable characteristics corresponding to each time window, and then inputting the transaction strategy applicable characteristics of each time window into a trained Bayesian classifier to obtain transaction strategy reference lines and early warning index values corresponding to each time window. In the transaction risk early warning method provided by the embodiment of the application, the transaction strategy feature prediction model is obtained based on long-time and short-time memory neural network training, the input of the transaction strategy prediction model is the change features of various transaction signals on a plurality of continuous time windows, and the feature change pattern of the transaction signals can be obtained by performing simulation analysis on the change of the transaction signal features on the continuous time windows, so that the influence of the change of the transaction signal features on the transaction strategy applicable features is expected, the incidence relation between the transaction signals of the fusion market and the transaction strategy applicable features in different time periods has predictability, and the reliability of transaction strategy risk early warning is greatly improved.

It is understood that the beneficial effects of the second aspect to the fifth aspect can be referred to the related description of the first aspect, and are not described herein again.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.

Fig. 1 is a schematic flow chart illustrating a transaction risk early warning method according to an embodiment of the present disclosure;

FIG. 2 is a schematic flow chart illustrating a process for obtaining time series data of a plurality of transaction signals according to an embodiment of the present application;

FIG. 3 is a schematic flow chart illustrating a process for obtaining a transaction policy feature prediction model according to an embodiment of the present application;

FIG. 4 is a schematic structural diagram of a first structural model provided in an embodiment of the present application;

FIG. 5 is a schematic flow chart of obtaining a first structural model according to an embodiment of the present application;

FIG. 6 is a schematic diagram of an analysis factor integrated clustering provided in an embodiment of the present application;

FIG. 7 is a flowchart illustrating a method for obtaining a Bayesian classifier according to an embodiment of the present application;

fig. 8 is a schematic structural diagram of a transaction risk early warning device according to an embodiment of the present application;

fig. 9 is a schematic structural diagram of a transaction risk early warning device according to another embodiment of the present application;

fig. 10 is a schematic structural diagram of a transaction risk early warning device according to an embodiment of the present application.

Detailed Description

In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather "one or more but not all embodiments" unless specifically stated otherwise. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless expressly specified otherwise.

The financial transaction risk early warning means that under the current transaction signal, early warning is provided for risks which may appear after the transaction strategy is adopted and implemented by a decision maker.

The financial market, the trading strategy and the decision maker are taken as research objects, and the trading information of the historical financial market is analyzed, so that available trading signals in the financial market are sparse, large loss is easily caused when the trading is carried out according to the trading signals blindly, and risk early warning needs to be given to frequent trading; different decision makers may have preferences on different transaction strategies, and the risk of the transaction strategy adopted by the decision maker needs to be early warned so as to restrict the decision openness of the decision maker; moreover, financial markets change infrequently, each trading strategy has its application range, and trading strategy applicability risk early warning needs to be given according to changes of market conditions.

Based on the above, the application provides a transaction risk early warning method, and aims to provide risk early warning information with high reliability.

The following describes the technical solutions of the present application and how to solve the above technical problems with specific embodiments. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below with reference to the accompanying drawings.

Fig. 1 is a schematic flow diagram of a transaction risk early warning method according to an embodiment of the present application, and as shown in fig. 1, the transaction risk early warning method includes:

s101, acquiring time series data of various transaction signals; the time series data includes characteristics of each transaction signal over a plurality of successive time windows.

In this embodiment, the transaction signal is a characteristic signal for characterizing a financial market change. For example, assuming that the financial market is analyzed based on a random index (also known as the KDJ index), the KDJ index consists of three curves, denoted by index line K, index line D, and index line J. Wherein, the K line is a quick confirmation line, when the value of the device is more than 90, the stock price is over-bought, and when the value is less than 10, the stock price is over-sold; d line is a slow speed trunk line, the numerical value of the D line is more than 70, which indicates that the stock price is overburdened, and the numerical value is less than 30, which indicates that the stock price is overburdened; and when the value of the indicator line J is more than 100, the stock price is in overbilling, and when the value of the indicator line J is less than 0, the stock price is overbilling. The three lines are crossed with each other, so that a proper buying point can be found from the crossed position, and the trading signal can be the crossed point of the K line and the D line, the inflection point of the D line and the like.

The financial market may be a trading market for securities, futures, and spot among others.

There are a variety of trading signals for financial markets. For example: the trading signals may be classified into approach signals, level signals and stop-loss signals according to the trading strategy that should be adopted. The entrance signal refers to starting to build a bin when the price difference deviates from a first preset standard difference, the leveling signal refers to leveling when the price difference returns to zero, and the loss stopping signal refers to stopping loss when the price difference cannot return to zero but continues to expand to a second standard difference. It is understood that the second standard deviation is greater than the first standard deviation.

Since the transaction signals are dynamic, in this embodiment, changes in the financial market are characterized by time series data of the various transaction signals. Acquiring time-series data of the plurality of transaction signals comprises acquiring time-series data of each transaction signal in the plurality of transaction signals; the time series data for each transaction signal includes characteristics of the transaction signal in each of successive time windows. The time window is a time period, and the length of the time window can be preset; for example, the length of the time window may be set to one day.

In practical application, transaction signals in the financial market are sparse, and are specifically characterized in that the transaction signals are not present in part of time windows on a plurality of continuous time windows in the same transaction signal. In order to obtain the characteristics of each transaction signal in each of the successive time windows, in one embodiment, the characteristic change of the transaction signal can be tracked to obtain the characteristics of the transaction signal in the next adjacent time window according to the characteristics of the transaction signal in the current time window.

The manner in which the characteristic change of the transaction signal is tracked and the transaction signal of the next adjacent time window is obtained is exemplarily described in the embodiment shown in fig. 2.

S102, inputting the time sequence data into a transaction strategy feature prediction model to obtain applicable features of the transaction strategy corresponding to each time window; the transaction strategy feature prediction model is obtained based on long-time and short-time memory neural network training.

In the present embodiment, the trading strategy application feature is used to describe the application scope and application situation of the trading strategy. There are a variety of trading strategies for financial markets, such as: the trading strategy includes, but is not limited to, buy, sell, etc. operations. Different trading strategies have different application situations, namely when different trading signals appear in the financial market, the applicable trading strategies are different.

In this embodiment, the transaction strategy prediction model is obtained based on long-and-short-term memory neural network training. A Long Short-Term Memory neural network (LSTM) is a time recurrent neural network, and is used to predict events with Long intervals and delays in a time sequence. According to the embodiment, the LSTM is used for modeling the time series data of the trading signals in a long time domain according to the size of a preset time window, so that the association relation between the trading signals of the melting market and expected market behaviors in different time periods has predictability; for example, a transaction signal of a next time window may be expected based on a current transaction signal, and a corresponding transaction policy applicable characteristic may be obtained according to the transaction signal prediction of the next time window, so as to perform risk early warning according to the transaction policy applicable characteristic. Where the expected market behavior is characterized by trading strategy applicability.

In this embodiment, the sample time series of various transaction signals and the applicable characteristics of the transaction strategy corresponding to the sample time series are used as a training sample; and obtaining a plurality of training samples of the target financial market in a historical period, and training the long-time memory neural network and the short-time memory neural network through the plurality of training samples to obtain a trading strategy characteristic prediction model. Wherein the plurality of trading signals are all kinds of trading signals that can exist in the target financial market, and the time sequence of each kind of trading signal comprises the characteristics of the trading signal in continuous time windows of historical periods.

In this embodiment, inputting the time-series data into the transaction policy feature prediction model to obtain the applicable transaction policy features corresponding to the time windows includes sequentially inputting the time-series data of the transaction signals into the transaction policy feature prediction model according to the time sequence of a plurality of continuous time windows to obtain the applicable transaction policy features of each time window, and storing and outputting the applicable transaction policy features of each time window.

The applicable characteristics of the trading strategy of a plurality of adjacent or continuous time windows can be the same, namely the financial market is not changed greatly in a longer period of time, and trading can be carried out based on the same trading strategy.

S103, inputting the applicable characteristics of the transaction strategy of each time window into a trained Bayesian classifier, and obtaining a transaction strategy reference line and an early warning index value corresponding to each time window.

In this embodiment, after the applicable characteristics of the transaction strategy of each time window are determined, the transaction strategy reference line and the early warning index value of each time window can be obtained by using the trained bayesian classifier.

The Bayesian classifier is a probability classifier based on Bayesian theorem.

For example, the principle of the bayesian classifier can be referred to as follows:

let X { a1, a2, … Am } be an item to be classified, where a1-Am are the characteristic attributes of X.

The category set C is { Y1, Y2 … Yn }, and is used to indicate that X is divided into several categories, and n is the number of categories.

The calculation yields P (Y1| X), P (Y2| X), … P (Yn | X), where P is the probability.

If P (Yk | X) ═ max { P (Y1| X), P (Y2| X), … P (Yn | X) }, X ∈ Yk, classification of X is achieved.

In this embodiment, X is a transaction policy applicable characteristic, and Yn is a classification of different transaction policy reference lines and early warning index values. Various trading strategy reference lines and early warning index values of the trading strategy reference lines can be obtained in advance according to financial market historical data.

In this embodiment, the bayesian classifier can be a gaussian naive classifier.

In this embodiment, the Bayesian classifier is trained in advance. Acquiring applicable characteristics of a plurality of transaction strategies, acquiring early warning index values of a plurality of transaction strategy reference lines and each transaction strategy reference line, and forming a classification training sample by using the applicable characteristics of each sample transaction strategy and the transaction strategy reference lines and the early warning index values corresponding to the applicable characteristics of the sample transaction strategies; and carrying out classification training on the Bayes classifier according to the plurality of classification training samples to obtain the trained Bayes classifier.

The transaction risk early warning method provided by the embodiment of the application obtains the change characteristics of various transaction signals in the financial market on a plurality of continuous time windows; and inputting the characteristics of various transaction signals which change on a plurality of continuous time windows into a transaction strategy characteristic prediction model to obtain transaction strategy applicable characteristics corresponding to each time window, and then inputting the transaction strategy applicable characteristics of each time window into a trained Bayesian classifier to obtain transaction strategy reference lines and early warning index values corresponding to each time window. In the transaction risk early warning method provided by the embodiment of the application, the transaction strategy feature prediction model is obtained based on long-time and short-time memory neural network training, the input of the transaction strategy prediction model is the change features of various transaction signals on a plurality of continuous time windows, and the feature change pattern of the transaction signals can be obtained by performing simulation analysis on the change of the transaction signal features on the continuous time windows, so that the influence of the change of the transaction signal features on the transaction strategy applicable features is expected, the incidence relation between the transaction signals of the fusion market and the transaction strategy applicable features in different time periods has predictability, and the reliability of transaction strategy risk early warning is further improved.

In practical applications, the change of the transaction signal may not be continuous, and in order to obtain the change characteristics of the transaction signal in the continuous time window, the change of the characteristics of the transaction signal may be tracked to obtain the characteristics of the transaction signal in the next adjacent time window according to the characteristics of the transaction signal in the current window. This is illustrated by the embodiment shown in fig. 2.

Fig. 2 is a schematic flow chart illustrating a process of acquiring time-series data of a plurality of transaction signals according to an embodiment of the present application. The embodiment of the present invention is based on the embodiment of fig. 1, and the step S101 of obtaining time-series data of various transaction signals is exemplarily described. As shown in fig. 2, time series data of a plurality of transaction signals is obtained, including:

s201, acquiring a transaction signal set of each time window in a plurality of continuous time windows; each set of transaction signals includes a feature vector for a plurality of types of transaction signals, the feature vector for each type of transaction signal including feature values for a plurality of analysis factors associated with the transaction signal.

In this embodiment, the transaction signal may be characterized by an analysis factor. The analysis factors are all factors which can cause the financial market to generate target trading signals; for example, the analysis factors may include government policies, macro-economic conditions, historical financial status of each listed company, competitive advantage of each listed company, and hot news, among others. Analysis factors related to each transaction signal can be obtained through industry market analysis, company research and the like; it should be understood that the analysis factor may be different for each transaction signal.

S202, comparing the transaction signal sets of two adjacent time windows in the multiple continuous time windows to obtain multiple groups of association vectors between the two adjacent time windows.

In this embodiment, comparing the transaction signal sets of two adjacent time windows in the multiple continuous time windows refers to comparing the transaction signal sets of all two adjacent time windows in the multiple continuous time windows.

For example, assuming that the multiple continuous time windows are 5 continuous time windows numbered 1-5, comparing the transaction signal sets of two adjacent time windows in the multiple continuous time windows means comparing the transaction signal sets of time windows numbered 1 and 2, 2 and 3, 3 and 4, and 4 and 5 to obtain 4 sets of association vectors.

In this embodiment, each set of associated vectors is two eigenvectors with the most similar eigenvalues in adjacent time windows; the two feature vectors belong to different time windows of two adjacent time windows respectively.

Each feature vector comprises feature values of a plurality of analysis factors, and the two feature vectors with the most similar feature values refer to the two feature vectors with the most similar analysis factors and the same feature values.

For example, assume transaction signal set W for time window number 1₁And the transaction signal set of No. 2 time window is W₂And the transaction signal set of No. 3 time window is W₃(ii) a Assume that each time window includes k eigenvectorsThe feature vectors in the transaction signal set W1 are a1, a2₂K eigenvectors of (a) are respectively b1, b2, … bk, and a transaction signal set W₃The k eigenvectors of (a) are c1, c2, … ck, respectively. Firstly, W is₁The feature vector of (1) and W₂The feature vectors in (1) are compared one by one, if W is₁In (a) 1 and W₂Where b2 has the most common eigenvalue, then W₁In (a) 1 and W₂B2 in (1) constitutes an association vector. Further, transaction signal set W of time window No. 2 is divided into₂Transaction signal set W with time window No. 3₃Making a comparison, if W₂B2 and W₃Where c4 has the most common eigenvalue, then W₂B2 and W₃C4 in (1) constitutes an association vector; and by analogy, obtaining the association vector between all the two adjacent time windows.

S203, for each feature vector, sequentially connecting all the associated vectors of the feature vector in a plurality of continuous time windows according to a time sequence, generating a feature curve of the transaction signal corresponding to the feature vector, and obtaining a transaction signal change feature corresponding to the feature curve based on a migration model.

In this embodiment, for each feature vector, all the associated vectors of the feature vector in a plurality of consecutive time windows are sequentially connected in time order to generate a feature curve of the transaction signal corresponding to the feature vector, and a feature value of the transaction signal in the next time window can be predicted according to the feature curve.

And acquiring the transaction signal change characteristics corresponding to the characteristic curve based on the migration model, including acquiring the change characteristics of the transaction signal between two adjacent time windows based on the migration model, and acquiring the transaction signal change modes corresponding to the characteristic curve according to the change characteristics.

The transaction signal variation pattern may be a variation and a trend of the transaction signal within a time window.

According to the transaction risk early warning method provided by the embodiment of the application, the characteristic change of the transaction signal is tracked, the characteristic of the transaction signal of the next adjacent time window can be obtained according to the transaction signal characteristic of the current window, and the change characteristic of the transaction signal on the continuous time window is further obtained.

Fig. 3 is a schematic flow chart of obtaining a transaction policy feature prediction model according to an embodiment of the present application, and illustrates one possible embodiment of obtaining the transaction policy feature prediction model before inputting time series data into the transaction policy feature prediction model to obtain applicable transaction policy features corresponding to each time window in step 102. As shown in fig. 3, the process of obtaining the transaction policy feature prediction model includes:

s301, obtaining a plurality of analysis factors related to the transaction signals and a plurality of transaction strategies of the target decision maker.

In this embodiment, the analysis factors are all factors that may cause the financial market to generate a target trading signal. There may be more than one analysis factor for each transaction signal. Analysis factors include, but are not limited to, government policies, macro-economic situation, historical financial status of each listed company, competitive advantage of each listed company, and hot news, among others. It should be understood that there is a correlation between different analytical factors.

The analysis factor of each transaction signal can be obtained through industry market analysis, company investigation and the like.

The trading strategy to be employed by different decision makers may vary. For example, when the roles of the decision makers are different, the transaction strategies adopted by the decision makers may be different for the same transaction signals.

Due to the role of the decision maker and the preference of the decision maker, the method is open. In this embodiment, only a plurality of transaction policies of a certain target decision maker are obtained. When the target decision maker changes, the trading strategy of the decision maker changes.

S302, respectively obtaining a first structural model between the transaction signals and the analysis factors and a second structural model between the transaction strategies and the analysis factors based on a complete interpretation structure modeling method.

A Total Indirect Structure Model (TISM) is completely explained, a complex system is decomposed into a plurality of subsystem elements, and finally a multi-level hierarchical Structure Model is formed, wherein the Structure Model uses a directed graph to describe the association relationship between analysis factors of the system and the influence of the analysis factors on a trading signal/trading strategy.

The directed graph is a modeling description method of the connection relation among nodes and comprises the nodes and directed edges connecting the nodes. Nodes with influence relation in the directed graph are connected through directed edges, and the nodes at the input end of the directed edges have influence on the nodes at the output end of the directed edges; when there is no directed edge between two nodes, it means that the two nodes do not affect each other.

In this embodiment, each analysis factor is a node of a directed graph, any two analysis factors having an association relationship are connected by a directed edge, the directed edge is a straight line with an arrow, and the arrow of the directed edge points to the affected analysis factor.

For example, assume that a transaction signal is a D-line over K-line. The analytical factors for the transaction signal include government policy, macro economic situation, related stock market, historical financial status of listed company a, financial forecast of listed company a, competitive advantage of listed company a, margin of safety, and usefulness of the transaction signal. A first structural pattern between the transaction signal and the analysis factor is shown in fig. 4. In fig. 4, the straight line with an arrow is a directed edge, and the arrow direction of the directed edge is the output end of the directed edge. For example: the macro-economic situation and the government policy are connected through the directed edges, and the arrows of the directed edges point to the government policy, so that the macro-economic situation is characterized to have influence on the government policy.

In this embodiment, the method of acquiring the first structure interpretation model and the second structure interpretation model is the same. The manner in which the first/second structural models are obtained by the TISM-based method will be exemplarily described below.

S303, obtaining the use will of the target decision maker on the multiple transaction strategies, and generating expected confirmation models of the multiple transaction strategies according to the second structure model and the use will.

In order to solve the problem of great economic loss caused by the fact that decision makers make decisions by experience and some transaction strategies are preferred (use will) in the prior art, the preferences of target decision makers for a plurality of transaction strategies need to be included in analysis so as to restrict decision openness of the decision makers and prompt early warning.

In the embodiment, the use willingness of the target decision maker to the plurality of transaction strategies is obtained based on the expectation-confirmation theory.

The expectation-validation theory considers that the expectation-validation of a decision-maker after using a certain transaction strategy affects its perceived usefulness and satisfaction with the transaction strategy, which in turn affects the decision-maker's will to continue using the transaction strategy in the future. Wherein the expected confirmation is a comparison difference between the perceived effect of the target decision-maker after using a certain trading strategy and the initial expected effect.

Illustratively, generating a desired confirmation model of the plurality of transaction policies according to the second structural model and the willingness-to-use includes the steps of:

the following assumptions are first proposed:

h1: the desired degree of certainty of a decision maker has a significant positive impact on perceptual usefulness;

h2: the desired degree of confirmation of the decision maker has a significant positive impact on satisfaction;

h3: the perceptual usefulness of the decision-maker has a significant positive impact on usage satisfaction;

h4: the perception usefulness of the decision maker has a significant positive impact on the willingness to continue use;

h5: the satisfaction of the decision maker has a significant positive impact on the willingness to continue use.

Then taking the second structure model as input, constructing a hypothesis theoretical model based on the hypothesis, analyzing based on a structure equation model to check the hypothesis theoretical model, and checking the fitting degree condition of the hypothesis theoretical model through a fitting degree index value obtained by calculation and a preset judgment standard; and checking whether each applicable hypothesis is established or not in sequence according to the fitting degree condition. And if so, taking the current hypothesis theoretical model as an expected confirmation model, and determining the characteristic value of the current analysis factor in the second structure model as the applicable characteristic of the trading strategy.

And S304, generating a long-term memory neural network according to the first structural model and the expected confirmation model.

In this embodiment, the transaction policy feature prediction model includes an input layer, a hidden layer, and an output layer. Wherein the number of neurons in the input layer is the same as the number of the plurality of transaction signals in the first structural model. The number of the neurons of the output layer is the same as the number of the transaction strategies in the second structural model.

Generating a long-term and short-term memory neural network according to the first structural model and the expected confirmation model, and the method comprises the following steps: forming the neural network nodes corresponding to the transaction signals in the first structural model into an input layer, and forming the neural network nodes corresponding to the transaction strategies in the expected model into an output layer; and connecting the neural network nodes corresponding to the analysis factors according to the connection relationship between the analysis factors in the first structural model and the connection relationship between the analysis factors in the expected confirmation model, so as to generate the hidden layer of the long-time memory neural network.

S305, obtaining a plurality of training samples; each training sample comprises sample time sequence data of various transaction signals and applicable characteristics of a transaction strategy corresponding to the sample time sequence data.

The sample time series data of the transaction signals in the training samples are characteristic of changes in the transaction signals over successive time windows of a plurality of historical times.

S306, training the long-time and short-time memory neural network according to the plurality of training samples to obtain a transaction strategy characteristic prediction model.

In this embodiment, the training of the long-term and short-term memory neural network uses a back propagation algorithm. Illustratively, the output value of each neuron is calculated forward first, and then the error term value of each neuron is calculated backward, and the gradient of each weight is calculated according to the corresponding error term. Wherein the back propagation of the LSTM error term may include two directions: one is the backward propagation along the time, namely, the error term of each moment is calculated from the current t moment; one is to propagate the error term up one layer.

The transaction risk early warning method provided by the embodiment of the application enables the preference of a target decision maker on a plurality of transaction strategies to be brought into analysis, so that the decision openness of the decision maker is restrained, early warning is prompted, and the reliability of risk early warning is improved.

Fig. 5 is a schematic flow chart illustrating a process of obtaining a first structural model according to an embodiment of the present application. The present embodiment exemplifies how to obtain the first structural model in step S301 on the basis of the embodiment of fig. 3. As shown in fig. 5, the method for obtaining a first structural model between a plurality of transaction signals and a plurality of analysis factors based on a fully interpreted structural modeling method includes:

s501, clustering the plurality of molecular factors based on an integrated clustering algorithm to obtain a plurality of analysis factor types.

In this embodiment, the analysis factor is obtained by way of industry market analysis, company research on marketing, and the like.

The transaction signal is described through the molecular factors, and in order to solve the problem that the transaction signal description dimension is too high, in this embodiment, a plurality of analysis factors are clustered based on an inheritance clustering algorithm, so that a plurality of analysis factor types are obtained.

In this embodiment, the analysis factors may be clustered using a hierarchical binary K-means algorithm. Assuming that the directly obtained analysis factor is used as an initial analysis factor, the initial factor is taken as a root node, the analysis factor is divided into two nodes by using a standard K-means algorithm, each node is evaluated according to different transaction signal characteristics and requirements, and if the node is a leaf node, the node is not divided into two child nodes. Otherwise, the node is an internal node, and the node is divided into two parts by using the standard K-mean algorithm continuously until the generated nodes are all leaf nodes. Wherein, the evaluation index is the similarity between the nodes.

For example, please refer to fig. 6 together, and fig. 6 is a schematic diagram of an analysis factor integrated cluster according to an embodiment of the present application. As shown in fig. 6, w is a plurality of analysis factors obtained initially, and is considered as a root node. C1-C10 are child nodes, and each child node is a component cluster of a certain transaction signal. In the clustering process, component clusters comprising a plurality of analysis factors are obtained by changing the input characteristics and requirements of a transaction signal, then the similarity between different component clusters is judged by using preset cluster evaluation indexes, and the clustering result with the highest similarity with other component clusters is selected to obtain a plurality of analysis factors corresponding to the transaction signal.

S502, fitting the analysis factors respectively based on a genetic algorithm to obtain the fitting degree of each analysis factor, and taking the analysis factors with the fitting degree higher than a preset value as target analysis factors.

In order to improve the rationality and reliability of the analysis factor clustering results, the analysis factors may be fitted based on a genetic algorithm. The genetic algorithm is a randomized search method evolved by using the evolution law (survival of fittest and selection and elimination of genetic mechanism) of the biological world for reference.

In this embodiment, the analysis factors are regarded as genetic genes, the analysis factors are randomly selected from the gene pool to form a fitting model, and the genetic variation process is the genetic variation of the fitting model.

In this embodiment, the degree of fitting is the distribution of the analysis factors in the integrated clustering result; the more uniform the distribution, the more reasonable the fitting model is, and the higher the fitting degree is.

Illustratively, for the fitting model of the transaction signal, after one genetic variation, namely one iteration of the genetic algorithm, the goodness and badness of each fitting model are evaluated by the model fitting degree to serve as the basis for the elimination of the goodness and badness. Specifically, the distribution of the analysis factors in the transaction signal fitting model in the integrated clustering result is evaluated, and the more uniform the distribution is, the more reasonable the fitting model is. And (3) carrying out fitting degree evaluation on the fitting model generated in each generation, wherein the fitting model with high fitting degree has higher inheritance probability, and the model with poor fitting degree has higher mutation probability, so that the generation inherits. In order to prevent the number of fitting models from being severely expanded, when the number of fitting models reaches a preset number, fitting models with ranked fitting degrees are eliminated first.

S503, acquiring a first structural model between a plurality of transaction signal combinations and a plurality of target analysis factors based on a complete interpretation structure modeling method; the first structural model is used for representing the association relation between a plurality of transaction signals and a plurality of target analysis factors.

In this embodiment, reference may be made to steps 5031 to 5035 in the method for obtaining the first structural model.

5031. Establishing an adjacency matrix between the transaction signals and the target analysis factors; the adjacent matrix is used for representing direct influence relations between different transaction signals and target analysis factors, if one transaction signal is related to one target analysis factor, 1 is filled in the corresponding position of the matrix, and if not, 0 is filled in the corresponding position of the matrix. It should be understood that the default index may also have an effect on itself.

5032. And calculating to generate a reachable matrix according to the adjacent matrix. The adjacency matrix may be converted into a two-dimensional matrix, with 0 s and 1 s replacing V, A, X, O. The specific calculation process is the prior art and is not described herein again.

5033. Carrying out grade division on the reachable matrix, removing the override relation, and establishing a structural model;

and calculating an accessible set, a leading set and a common set according to the accessible matrix, wherein when the common set is the same as the accessible set, namely the first level, the row and the column where the first level is located are removed from the accessible matrix, the next level of target analysis factors are continuously searched from the rest matrixes, and the like until the levels of all the target analysis factors are obtained.

5034. Drawing an explanation structure model diagram according to the structure model, and obtaining a two-dimensional interaction matrix from the explanation structure model diagram.

5035. And obtaining a fully explained structure model, namely the first structure model, according to the two-dimensional interaction matrix.

The transaction signals are the representation of the big data characteristics of the financial market, the transaction risk early warning method provided by the embodiment of the application constructs a TISM model between the transaction signals and the analysis factors, determines the incidence relation between all the transaction signals and the analysis factors, finds out each transaction signal and the related analysis factor thereof, and realizes the representation of the transaction signals through the analysis factors. Correspondingly, the incidence relation between all transaction strategies and the analysis factors can be determined by constructing a TISM model between the transaction strategies and the analysis factors, and then the incidence relation between the transaction signals and the transaction strategies can be established based on the analysis factors.

On the other hand, in the embodiment, before the TISM model between the transaction signal and the analysis factor is constructed, the analysis factor is preprocessed based on the integrated clustering and genetic algorithm, so that the rationality and reliability of the clustering result of the analysis factor are effectively improved, and the rationality of the TISM model is further improved.

The technical scheme for obtaining the second structural model between the multiple transaction strategies and the multiple analysis factors based on the fully explained structural modeling method is the same as that in the embodiment shown in fig. 5, and only the transaction signal needs to be changed into the transaction strategy, which is not described herein again.

Fig. 7 is a schematic flow chart of obtaining a bayesian classifier according to an embodiment of the present application, which illustrates a possible implementation manner of obtaining the bayesian classifier before inputting the applicable transaction strategy features of each time window into the trained bayesian classifier and obtaining the transaction strategy reference lines and the early warning index values corresponding to each time window before step 103 in the embodiment shown in fig. 1. Referring to fig. 7, the method for obtaining the bayesian classifier comprises:

s701, obtaining applicable characteristics of sample transaction strategies of a plurality of historical continuous time windows.

The sample trading strategy application characteristics are used for describing the application range and the application condition of the sample trading strategy. When different trading signals appear in the financial market, the applicable trading strategies are different. For example, assuming that the financial market is analyzed based on the KDJ criteria, the applicable characteristics of the trading strategy may be a D-line drop 20, a D-line re-break 20, etc.

The length of the time windows of the plurality of consecutive time windows in this embodiment is the same as the length of the time windows in step 101.

S702, obtaining the decision preference of a target decision maker, and determining a transaction strategy reference line and an early warning index value of the applicable characteristics of each sample transaction strategy according to the decision preference.

The goal-decider in this embodiment is the same as the goal-decider in the embodiment shown in fig. 3.

The decision preferences of the decision-maker include the willingness of the decision-maker to use different transaction policies.

In this embodiment, determining the transaction strategy reference lines and the early warning index values of the applicable features of each sample transaction strategy according to the decision preference includes first obtaining a plurality of transaction strategy reference lines and early warning index values of different time windows, then classifying the transaction strategy reference lines and the early warning index values, and connecting the applicable features of each sample transaction strategy with the transaction strategy reference lines and the early warning index values of different categories according to a preset rule. Wherein the preset rules are determined by the decision preferences of the target decision maker.

Illustratively, when the decision preference of the objective decision maker is risk type, compared with a robust objective decision maker, the applicable characteristics of the same sample transaction strategy can reduce the benchmark criminal and early warning index values of the transaction strategy.

And S703, forming a classification training sample by the applicable characteristics of each sample transaction strategy, the transaction strategy reference line corresponding to the applicable characteristics of the sample transaction strategy and the early warning index value.

And S704, carrying out classification training on the Bayes classifier according to the plurality of classification training samples to obtain the trained Bayes classifier.

In this embodiment, the transaction strategy reference line and the early warning index value are classified in advance. The process of performing classification training on the bayesian classifier according to the plurality of classification training samples is similar to the training process in the prior art, and is not repeated herein.

According to the transaction risk early warning method provided by the embodiment of the application, when the classification training samples are obtained, the influence of decision preference on the transaction strategy reference line and the early warning index value is considered, namely the transaction strategy application characteristics are applied to the same sample, decision preference of decision makers is different, the corresponding transaction strategy reference line and the corresponding early warning index value can be different, the transaction strategy reference line and the corresponding early warning index value are adaptively adjusted, and the sensitivity and the reliability of the transaction side Kurl benchmark criminal and the early warning index value are improved.

It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

Based on the transaction risk early warning provided by the embodiment, the embodiment of the invention further provides an embodiment of a device for implementing the method.

Fig. 8 is a schematic structural diagram of a transaction risk early warning device according to an embodiment of the present disclosure. As shown in fig. 8, the transaction risk early warning device 80 includes: an acquisition module 801, a prediction module 802, and an early warning module 803.

An obtaining module 801, configured to obtain time-series data of multiple transaction signals; the time series data comprises a variation characteristic of each transaction signal over a plurality of successive time windows;

the prediction module 802 is configured to input the time series data into a transaction policy feature prediction model to obtain applicable transaction policy features corresponding to each time window; the transaction strategy feature prediction model is obtained based on long-time and short-time memory neural network training;

and the early warning module 803 is configured to input the applicable transaction strategy characteristics of each time window into the trained bayesian classifier, and obtain a transaction strategy reference line and an early warning index value corresponding to each time window.

An obtaining module 801, configured to obtain a transaction signal set of each time window in a plurality of consecutive time windows; each transaction signal set comprises a plurality of transaction signal feature vectors, and each transaction signal feature vector comprises a plurality of analysis factor feature values related to the transaction signal;

comparing the transaction signal sets of two adjacent time windows in the plurality of continuous time windows to obtain a plurality of groups of association vectors between the two adjacent time windows; the association vectors are two eigenvectors with the most similar eigenvalues in two adjacent time windows;

and for each feature vector, sequentially connecting all the associated vectors of the feature vector in a plurality of continuous time windows according to a time sequence to generate a feature curve of the transaction signal corresponding to the feature vector, and obtaining the transaction signal change feature corresponding to the feature curve based on a migration model.

The prediction module 802 is specifically configured to sequentially input the time series data of the multiple transaction signals into the transaction policy feature prediction model according to the time sequence of multiple continuous time windows, obtain the transaction policy applicable feature of each time window, and store and output the transaction policy applicable feature of each time window.

The transaction risk early warning device provided by the embodiment of the application acquires the change characteristics of various transaction signals in a financial market on a plurality of continuous time windows; and inputting the characteristics of various transaction signals which change on a plurality of continuous time windows into a transaction strategy characteristic prediction model to obtain transaction strategy applicable characteristics corresponding to each time window, and then inputting the transaction strategy applicable characteristics of each time window into a trained Bayesian classifier to obtain transaction strategy reference lines and early warning index values corresponding to each time window. In the transaction risk early warning method provided by the embodiment of the application, the transaction strategy feature prediction model is obtained based on long-time and short-time memory neural network training, the input of the transaction strategy prediction model is the change features of various transaction signals on a plurality of continuous time windows, and the feature change pattern of the transaction signals can be obtained by performing simulation analysis on the change of the transaction signal features on the continuous time windows, so that the influence of the change of the transaction signal features on the transaction strategy applicable features is expected, the incidence relation between the transaction signals of the fusion market and the transaction strategy applicable features in different time periods has predictability, and the reliability of transaction strategy risk early warning is further improved.

Fig. 9 is a schematic structural diagram of a transaction wind direction warning device according to another embodiment of the present disclosure. As shown in fig. 9, the transaction risk early warning device 80 further includes a modeling module 804 and a training module 805. Wherein the content of the first and second substances,

a modeling module 804 for

Acquiring a plurality of analysis factors related to a plurality of transaction signals and a plurality of transaction strategies of a target decision maker;

respectively acquiring a first structural model between a plurality of transaction signals and a plurality of analysis factors and a second structural model between a plurality of transaction strategies and a plurality of analysis factors based on a fully explained structural modeling method;

obtaining the use will of the target decision maker on the plurality of transaction strategies, and generating an expected confirmation model of the plurality of transaction strategies according to the second structure model and the use will;

generating a long-term and short-term memory neural network according to the first structural model and the expected confirmation model;

obtaining a plurality of training samples; each training sample comprises sample time sequence data of various transaction signals and applicable characteristics of a transaction strategy corresponding to the sample time sequence data;

and training the long-time and short-time memory neural network according to the plurality of training samples to obtain a transaction strategy characteristic prediction model.

The modeling module 804 is specifically configured to:

clustering the plurality of molecular factors based on an integrated clustering algorithm to obtain a plurality of analysis factor types;

respectively fitting a plurality of analysis factors based on a genetic algorithm to obtain the fitting degree of each analysis factor, and taking the analysis factors with the fitting degree higher than a preset value as target analysis factors;

acquiring a first structural model between a plurality of transaction signal sets and a plurality of target analysis factors based on a fully explained structural modeling method; the first structural model is used for representing the association relation between a plurality of transaction signals and a plurality of target analysis factors.

The transaction strategy characteristic prediction model comprises an input layer, a hidden layer and an output layer; the modeling module 804 is further specifically configured to:

forming the neural network nodes corresponding to the transaction signals in the first structural model into an input layer, and forming the neural network nodes corresponding to the transaction strategies in the expected model into an output layer; and connecting the neural network nodes corresponding to the analysis factors according to the connection relationship between the analysis factors in the first structural model and the connection relationship between the analysis factors in the expected confirmation model, so as to generate the hidden layer of the long-time memory neural network.

Optionally, a training module 805 configured to:

obtaining sample transaction strategy applicable characteristics of a plurality of historical continuous time windows;

obtaining the decision preference of a target decision maker, and determining a transaction strategy reference line and an early warning index value of the applicable characteristics of each sample transaction strategy according to the decision preference;

forming a classification training sample by using the applicable characteristics of each sample transaction strategy, and a transaction strategy reference line and an early warning index value corresponding to the applicable characteristics of the sample transaction strategy;

and carrying out classification training on the Bayes classifier according to the plurality of classification training samples to obtain the trained Bayes classifier.

The transaction risk early warning device provided by the embodiment of the application enables the preference of a target decision maker on a plurality of transaction strategies to be brought into analysis, so that the decision openness of the decision maker is restrained, early warning is prompted, and the reliability of risk early warning is improved.

The transaction risk early warning device provided in the embodiments shown in fig. 8 and fig. 9 may be used to implement the technical solutions in the above method embodiments, and the implementation principle and the technical effect are similar, which are not described herein again.

Fig. 10 is a schematic diagram of a transaction risk early warning device according to an embodiment of the present application. As shown in fig. 9, the transaction risk early warning apparatus 100 of this embodiment includes: at least one processor 1001, a memory 1002, and computer programs stored in the memory 1002 and executable on the processor 1001. The transaction risk early warning device further comprises a communication component 1003, wherein the processor 1001, the memory 1002 and the communication component 1003 are connected by a bus 1004.

When the processor 1001 executes the computer program, the steps in the above embodiments of the transaction risk early warning method are implemented, for example, steps S101 to S103 in the embodiment shown in fig. 1. Alternatively, the processor 1001, when executing the computer program, implements the functions of the modules/units in the above-described device embodiments, for example, the functions of the modules 801 to 803 shown in fig. 8.

Illustratively, the computer program may be partitioned into one or more modules/units that are stored in the memory 1002 and executed by the processor 1001 to accomplish the present application. The one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution process of the computer program in the transaction risk early warning device 100.

It will be understood by those skilled in the art that fig. 10 is merely an example of a transaction risk early warning apparatus, and does not constitute a limitation to the transaction risk early warning apparatus, and may include more or less components than those shown, or combine some components, or different components, such as an input-output device, a network access device, a bus, etc.

The Processor 1001 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory 1002 may be an internal storage unit of the transaction risk early warning device, or may be an external storage device of the transaction risk early warning device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like. The memory 1002 is used for storing the computer programs and other programs and data required by the transaction risk early warning device. The memory 1002 may also be used to temporarily store data that has been output or is to be output.

The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended ISA (EISA) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, the buses in the figures of the present application are not limited to only one bus or one type of bus.

The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.

The embodiments of the present application provide a computer program product, which when running on a mobile terminal, enables the mobile terminal to implement the steps in the above method embodiments when executed.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can implement the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include at least: any entity or device capable of carrying computer program code to a photographing apparatus/terminal apparatus, a recording medium, computer Memory, Read-Only Memory (ROM), random-access Memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In certain jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and patent practice.

In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.

In the embodiments provided in the present application, it should be understood that the disclosed apparatus/network device and method may be implemented in other ways. For example, the above-described apparatus/network device embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implementing, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.

Claims

1. A transaction risk early warning method is characterized by comprising the following steps:

acquiring time sequence data of various transaction signals; the time series data comprises varying characteristics of each transaction signal over a plurality of successive time windows;

2. The transaction risk early warning method according to claim 1, wherein the obtaining of the time-series data of the plurality of transaction signals comprises:

obtaining a set of transaction signals for each of the plurality of consecutive time windows; each transaction signal set comprises a feature vector of the plurality of transaction signals, and the feature vector of each transaction signal comprises feature values of a plurality of analysis factors related to the transaction signal;

and for each feature vector, sequentially connecting all the associated vectors of the feature vector in the plurality of continuous time windows according to a time sequence to generate a feature curve of the transaction signal corresponding to the feature vector, and obtaining the transaction signal change feature corresponding to the feature curve based on a migration model.

3. The transaction risk early warning method of claim 1, wherein the inputting the time series data into a transaction policy feature prediction model to obtain applicable transaction policy features corresponding to each time window comprises:

and sequentially inputting the time series data of various transaction signals into the transaction strategy characteristic prediction model according to the time sequence of the continuous time windows, obtaining the transaction strategy applicable characteristic of each time window, and storing and outputting the transaction strategy applicable characteristic of each time window.

4. The transaction risk early warning method according to any one of claims 1 to 3, wherein before inputting the time series data into a transaction policy feature prediction model and obtaining applicable transaction policy features corresponding to respective time windows, the method further comprises:

obtaining a plurality of analysis factors related to the plurality of transaction signals and a plurality of transaction strategies of a target decision maker;

respectively acquiring a first structural model between the transaction signals and the analysis factors and a second structural model between the transaction strategies and the analysis factors based on a fully-interpreted structural modeling method;

obtaining the use willingness of a target decision maker to the plurality of transaction strategies, and generating expected confirmation models of the plurality of transaction strategies according to the second structure model and the use willingness;

generating the long-time memory neural network according to the first structural model and the expected confirmation model;

obtaining a plurality of training samples; each training sample comprises sample time sequence data of the transaction signals and applicable characteristics of a transaction strategy corresponding to the sample time sequence data;

and training the long-time and short-time memory neural network according to the plurality of training samples to obtain the transaction strategy characteristic prediction model.

5. The transaction risk early warning method according to claim 4, wherein the obtaining a first structural model between the plurality of transaction signals and the plurality of analysis factors based on a fully interpreted structural modeling method comprises:

obtaining a first structural model between the plurality of transaction signal sets and the plurality of target analysis factors based on a fully interpreted structural modeling method; wherein the first structural model is used for characterizing the association relationship between the plurality of transaction signals and the plurality of target analysis factors.

6. The transaction risk early warning method according to claim 4, wherein the transaction strategy feature prediction model comprises an input layer, a hidden layer and an output layer;

generating the long-term and short-term memory neural network according to the first structural model and the expected confirmation model, wherein the generating of the long-term and short-term memory neural network comprises:

forming the input layer by using the neural network nodes corresponding to the transaction signals in the first structural model, and forming the output layer by using the neural network nodes corresponding to the transaction strategies in the expected model; and connecting the neural network nodes corresponding to the analysis factors according to the connection relationship between the analysis factors in the first structural model and the connection relationship between the analysis factors in the expected confirmation model, so as to generate the hidden layer of the long-term memory neural network.

7. The transaction risk early warning method according to claim 1, wherein before inputting the applicable transaction strategy features of each time window into the trained bayesian classifier and obtaining the transaction strategy reference lines and early warning index values corresponding to each time window, the method further comprises:

obtaining decision preference of a target decision maker, and determining a transaction strategy reference line and an early warning index value of applicable characteristics of each sample transaction strategy according to the decision preference;

8. A transaction risk early warning device, comprising:

the acquisition module is used for acquiring time sequence data of various transaction signals; the time series data comprises varying characteristics of each of the plurality of transaction signals over a plurality of successive time windows;

9. A transaction risk warning device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor when executing the computer program implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.