Financial risk prediction method and device of enterprise, equipment, medium and program product

Abstract

The invention discloses a financial risk prediction method, device and equipment of an enterprise, a medium and a program product, which are applied to the field of financial science and technology. The method comprises the steps that operation data and financial market environment data are acquired for feature selection and / or feature extraction, and original feature representation is obtained; inputting into a pre-trained target encoder, and obtaining a compact feature representation matched with the target enterprise; inputting the compact feature representation into a pre-trained target ensemble learning framework; and carrying out weighted summation on individual financial risk prediction results output for the current compact feature representation according to the plurality of basic learners in the frame through the target ensemble learning frame to obtain a financial risk prediction result. According to the embodiment of the invention, a complete enterprise financial risk prediction method system is constructed, multi-source information is integrated, the prediction model is enabled to better adapt to market changes, errors of single model prediction are reduced through integration of learners and weighted summation, the reliability of risk prediction is improved, and the risk of misjudgment is reduced.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, applicable to the field of financial technology, and in particular to a method, device, equipment, medium and program product for predicting financial risks of an enterprise. Background Art

[0002] Today's financial markets are facing increasingly complex and diverse risks from market fluctuations, competitive pressures, natural disasters and other aspects. Effective risk management is crucial to the long-term development of enterprises. It can help enterprises identify, assess and respond to risks, reduce negative impacts and seize potential opportunities.

[0003] However, when dealing with complex risks, traditional risk management methods lack effective comprehensive consideration of different types of risks. Faced with massive and complex data, traditional methods have difficulties in data analysis and building accurate models. When faced with a complex and changing market environment, the decisions made by traditional methods are not accurate enough and are difficult to adapt to changes.

[0004] Based on this, there is an urgent need for a new method to predict the financial risks of enterprises in order to achieve more accurate enterprise risk prediction. Summary of the invention

[0005] Based on this, the present invention provides a method, device, equipment, medium and program product for predicting the financial risks of an enterprise to solve the problem that traditional risk management methods have low prediction accuracy and cannot provide timely response decisions when predicting the financial risks of an enterprise.

[0006] In a first aspect, an embodiment of the present invention provides a method for predicting financial risks of an enterprise, the method comprising:

[0007] Obtain the current operating data of the target enterprise and the current financial market environment data for feature selection and / or feature extraction to obtain the current original feature representation corresponding to the target enterprise;

[0008] Input the current original feature representation into the pre-trained target encoder to obtain the current compact feature representation that matches the target enterprise;

[0009] Input the current compact feature representation into the pre-trained target ensemble learning framework;

[0010] Through the target ensemble learning framework, the individual financial risk prediction results output by multiple basic learners in the framework for the current compact feature representation are weighted summed to obtain the target financial risk prediction results matching the target enterprise, where different basic learners are constructed based on different reinforcement learning algorithms.

[0011] In a second aspect, an embodiment of the present invention provides a financial risk prediction device for an enterprise, the device comprising:

[0012] The original feature representation acquisition module is used to obtain the current operating data of the target enterprise and the current financial market environment data for feature selection and / or feature extraction to obtain the current original feature representation corresponding to the target enterprise;

[0013] A compact feature representation acquisition module is used to input the current original feature representation into a pre-trained target encoder to obtain the current compact feature representation matching the target enterprise;

[0014] A feature input module, used to input the current compact feature representation into a pre-trained target ensemble learning framework;

[0015] The prediction result acquisition module is used to perform weighted summation of individual financial risk prediction results output by multiple basic learners in the framework for the current compact feature representation through a target ensemble learning framework to obtain a target financial risk prediction result that matches the target enterprise, wherein different basic learners are constructed based on different reinforcement learning algorithms.

[0016] In a third aspect, an embodiment of the present invention provides an electronic device, the electronic device comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute a method for predicting financial risks of an enterprise as described in any embodiment of the present invention.

[0020] In a fourth aspect, a computer-readable storage medium is also provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement a method for predicting financial risks of an enterprise as described in any embodiment of the present invention when executed.

[0021] In a fifth aspect, a computer program product is also provided, the computer program product comprising a computer program, and the computer program, when executed by a processor, implements a method for predicting financial risks of an enterprise as described in any embodiment of the embodiments of the present invention.

[0022] The technical solution of the embodiment of the present invention starts with obtaining the original data, obtaining the original feature representation through feature processing, further optimizing the features through the encoder to obtain the compact feature representation, and then using multiple basic learners in the integrated learning framework for prediction. Finally, the prediction results of each learner are weighted and summed to obtain the final risk prediction result. The whole process together constitutes a complete enterprise financial risk prediction method system, which fully integrates multi-source information so that the prediction model can learn different decision-making strategies to better adapt to market changes. It also balances the advantages and limitations of different learners through the integration of multiple basic learners and weighted summation, reduces the error of single model prediction, significantly improves the reliability of overall risk prediction, and reduces the risk of misjudgment.

[0023] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 is a flow chart of a method for predicting financial risks of an enterprise provided according to the first embodiment of the present invention;

[0026] Figure 2 is a flow chart of another method for predicting financial risks of an enterprise provided according to the second embodiment of the present invention;

[0027] Figure 3 This is a schematic diagram of the structure of a financial risk prediction device for an enterprise provided according to Embodiment 3 of the present invention;

[0028] Figure 4 It is a structural schematic diagram of an electronic device of a method for predicting financial risks of an enterprise provided according to a fourth embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] Embodiment 1

[0032] Figure 1 This is a flowchart of a method for predicting the financial risk of an enterprise provided in the first embodiment of the present invention. This embodiment is applicable to the case of predicting the financial risk of an enterprise. The method can be executed by a financial risk prediction device of an enterprise. The device can be implemented in the form of hardware and / or software. The device can be configured in a big data financial service platform. Figure 1 As shown, the method includes:

[0033] S110, obtaining the current operation data and the current financial market environment data of the target enterprise for feature selection and / or feature extraction, and obtaining the current original feature representation corresponding to the target enterprise.

[0034] This step is the first step in financial risk forecasting, namely data collection and feature extraction. Specifically, the current operating data of the target enterprise can refer to financial statements, sales receipts, and cost data, etc., which are used to reflect the financial status and operating conditions of the enterprise; the current financial market environment data can be interest rates, exchange rates, and stock price indices, etc., to reflect the macroeconomic environment in which the target enterprise is located.

[0035] Feature selection mainly refers to selecting features that have an important impact on risk prediction from the original data, and feature extraction mainly refers to extracting new feature vectors that are more helpful for prediction from the original data through some means (such as mathematical transformation, statistical analysis, etc.). The processed data or feature vector is called the current original feature representation corresponding to the target enterprise.

[0036] S120: Input the current original feature representation into a pre-trained target encoder to obtain a current compact feature representation that matches the target enterprise.

[0037] As a machine learning model, the target encoder maps high-dimensional data (current original feature representation) to a low-dimensional compact latent space to obtain a more compact representation that better reflects the essential characteristics of the data, while reducing the complexity of subsequent calculations. The target encoder is pre-trained with a large amount of historical data, so it can accurately convert the current original feature representation into the current compact feature representation that matches the target enterprise.

[0038] S130, inputting the current compact feature representation into a pre-trained target ensemble learning framework.

[0039] The compact feature representation processed by the encoder is used as the input of the ensemble learning framework to provide better feature data for the final risk prediction. The target ensemble learning framework consists of multiple basic learners built on different reinforcement learning algorithms. It combines the results obtained from different angles and levels of analysis of multiple basic learners to improve the prediction accuracy, which is the core processing link of the risk prediction process.

[0040] S140. A target ensemble learning framework is used to perform weighted summation on individual financial risk prediction results output by multiple basic learners within the framework for the current compact feature representation, to obtain a target financial risk prediction result that matches the target enterprise, wherein different basic learners are constructed based on different reinforcement learning algorithms.

[0041] In the target ensemble learning framework, each base learner will output its own financial risk prediction results based on the current compact feature representation. These results may be different because each base learner may be built based on different reinforcement learning algorithms, and they have different understanding and analysis capabilities of data. In order to obtain the final risk prediction result, the target ensemble learning framework will perform a weighted summation of these individual prediction results. The weights of the weighted summation are determined during the training process. They reflect the contribution and accuracy of each base learner in the overall prediction. Through weighted summation, a target financial risk prediction result that matches the target enterprise can be obtained.

[0042] The embodiment of the present invention starts with acquiring raw data, obtains raw feature representation through feature processing, further optimizes features through an encoder to obtain compact feature representation, then uses multiple basic learners in an integrated learning framework to make predictions, and finally weights and sums the prediction results of each learner to obtain the final risk prediction result. The entire process together constitutes a complete enterprise financial risk prediction method system, which fully integrates multi-source information so that the prediction model can learn different decision-making strategies to better adapt to market changes, and balances the advantages and limitations of different learners through the integration of multiple basic learners and weighted summation, thereby reducing the error of single model prediction, significantly improving the reliability of overall risk prediction, and reducing the risk of misjudgment.

[0043] Optionally, before inputting the current original feature representation into the pre-trained target encoder, the following may also be included:

[0044] Collect historical financial risk data of each enterprise, historical operating data associated with historical financial risk data, and matching historical financial market environment data;

[0045] According to each historical operation data and the matching historical financial market environment data, the autoencoder model is trained, and the encoder in the trained autoencoder model is used as the target encoder;

[0046] According to the target encoder, the historical financial risk data of each enterprise, the historical operation data associated with the historical financial risk data, and the matching historical financial market environment data are processed to obtain a training sample set;

[0047] The training sample set is used to train each basic learner in the pre-built integrated learning framework, as well as the weighted weights of each basic learner when performing integrated prediction, to obtain the target integrated learning framework.

[0048] The historical financial risk data of each enterprise records the financial risk situations encountered by each enterprise in the past, and is the target or label that the model needs to predict; and the historical operating data associated with the historical financial risk data reflects the operating conditions of the enterprise before encountering financial risks, including various financial indicators, business data, etc., which are potentially related to the occurrence of financial risks.

[0049] After training, the target autoencoder can compress historical operating data and financial market environment data into low-dimensional and compact feature representations, and combine the above feature representations with the corresponding historical financial risk data (as labels) to form a training sample set, which contains input features and labels for training the integrated learning framework and is the basis for subsequent model training. The integrated learning framework is a model that contains multiple basic learners, which can be different types of machine learning algorithms, such as decision trees, support vector machines, or neural networks, and the weighted weights determine the contribution of each basic learner to the final prediction result.

[0050] In an embodiment of the present invention, the autoencoder model and the integrated learning framework are trained based on these rich data, so that the model can learn risk characteristics and laws from a large amount of actual data, ensuring that the model training has a solid data foundation, improving the model's ability to fit real risk situations, and thereby improving the model's prediction effect in practical applications.

[0051] Furthermore, the autoencoder model is trained based on the historical operation data and the matching historical financial market environment data, including:

[0052] Perform feature selection and / or feature extraction on each historical operation data and the matching historical financial market environment data to obtain multiple historical original feature representations;

[0053] Input each historical original feature representation into the pre-built autoencoder model respectively, and train the autoencoder model;

[0054] The autoencoder model includes an encoder composed of a multi-layer neural network and a decoder symmetrical to the encoder network structure. The autoencoder model uses an adaptive moment estimation algorithm to minimize the reconstruction error.

[0055] Before training the autoencoder, it is necessary to preprocess the historical operating data and financial market environment data, that is, to perform feature selection or feature extraction. After feature selection or feature extraction, we obtain multiple historical original feature representations, which will be used as the input of the autoencoder. Each historical original feature representation is passed as input to the pre-built autoencoder model. The autoencoder model continuously adjusts the model parameters by comparing the difference between the input (i.e., the historical original feature representation) and the output (i.e., the reconstructed feature representation) to minimize this difference (i.e., the reconstruction error).

[0056] An autoencoder is an unsupervised learning neural network that reconstructs input data through the structure of an encoder and a decoder. In the above process, the encoder is responsible for compressing the input data into a low-dimensional representation (i.e. encoding), while the decoder is responsible for reconstructing this low-dimensional representation back to the original data (or a form as close to the original data as possible).

[0057] In addition, the adaptive moment estimation algorithm (such as the Adam algorithm) is an adaptive learning rate method based on first-order and second-order moment estimation. During the training process, the adaptive moment estimation algorithm adjusts the learning rate according to the historical gradient information and the exponential decay average of the gradient square, thereby more effectively minimizing the reconstruction error. Through continuous iteration and optimization, the autoencoder model can learn an effective representation of the data, which can be used in subsequent risk prediction tasks.

[0058] In an embodiment of the present invention, the autoencoder model used adopts an encoder and decoder structure composed of a multi-layer neural network, and uses an adaptive moment estimation algorithm to minimize the reconstruction error, which helps the model to better learn the intrinsic structure and feature representation of the data, and improves the encoding and decoding capabilities of the autoencoder model for the data, thereby laying the foundation for the subsequent generation of high-quality compact feature representations.

[0059] Optionally, according to the target encoder, the historical financial risk data of each enterprise, the historical operation data associated with the historical financial risk data, and the matching historical financial market environment data are processed to obtain a training sample set, which may include:

[0060] Inputting each historical operation data respectively associated with each historical financial risk data and the matched historical financial market environment data into the target encoder respectively, and obtaining the historical compact feature representation corresponding to each historical financial risk data respectively;

[0061] Generate financial risk labels corresponding to each historical financial risk data, and combine each financial risk label with the matching historical compact feature representation to obtain a training sample set.

[0062] The historical operating data mainly refers to the operating data of the enterprise in different periods, and is associated with the historical financial risk data. The historical financial market environment data refers to the interest rates, exchange rates, stock indexes, etc. in different periods, which match the above historical operating data in terms of period. Financial risk labels are used to indicate the size of financial risks, and are usually generated based on known historical financial risk data. The specific form of the label can be a numerical value (such as risk score), a level (such as high risk, medium risk and low risk) or other identifiers that can indicate the size of the risk.

[0063] The financial risk label is associated or paired with the historical compact feature representation to form a complete data point, and a set containing multiple data points is the training sample set, so that the subsequent training model can learn the mapping relationship from operational data and financial market environment data to financial risk labels.

[0064] In an embodiment of the present invention, historical data is passed through a target encoder to obtain a historical compact feature representation, and is combined with the generated financial risk labels to form a training sample set. Such a training sample set is based on an optimized feature representation and contains clear risk label information, which can provide more accurate and targeted training data for the basic learner in the integrated learning framework, and help the basic learner to better learn the relationship between risk features and risk labels, thereby improving the prediction ability of the model; at the same time, since the samples in the training sample set are associated with clear financial risk labels, during the model training and prediction process, the decision-making process and results of the model can be explained and analyzed based on these labels, which is of great significance for enterprises to understand the reasons for the risk prediction results and explain risk management measures to stakeholders such as regulators and investors, and enhances the transparency and credibility of the risk management method.

[0065] Furthermore, before using the training sample set to train each basic learner in the pre-built integrated learning framework and the weighted weights of each basic learner when performing integrated prediction, the following may also be included:

[0066] Learning the training sample set through a self-supervised learning model, wherein the self-supervised learning model is a generative adversarial network, and the generative adversarial network includes a generative model and a discriminative model;

[0067] After learning is completed, generating samples are collected from the generating model in the self-supervised learning model, and the generating samples are screened and / or corrected;

[0068] The generated samples that have completed the screening and / or correction process are used as expansion samples and added to the training sample set.

[0069] Generative adversarial network is a deep learning model, which consists of two parts: generative model and discriminative model. The generative model is responsible for generating new samples that are as close to real data as possible, while the discriminative model is responsible for distinguishing the generated samples from the real samples. These two models continuously improve their respective performances through mutual competition and confrontation, and finally enable the generative model to generate very realistic data. In an embodiment of the present invention, under the framework of self-supervised learning, the generative adversarial network can be used to generate new samples similar to the training sample set, and these new samples can be used as expansion samples to enhance the generalization ability of the model.

[0070] Completion of learning means that the training process of the generative adversarial network has been completed, and both the generative model and the discriminative model have reached a relatively stable state. At this time, the generated samples can be collected from the generative model. Since the generative model may produce some samples of low quality, screening means that the samples with higher quality can be retained by a certain standard (such as similarity with the real samples), and correction can refer to the adjustment and optimization of the samples of low quality. The generated samples that have been screened and / or corrected can be regarded as expanded samples. These samples, together with the original training sample set, constitute a larger data set.

[0071] In the embodiment of the present invention, the training sample set is learned through the self-supervised learning model (generative adversarial network), and the generated samples are collected and added to the training sample set, which effectively expands the scale of the training data. At the same time, the generative adversarial network can learn the distribution characteristics of the data and generate samples with a certain diversity, which helps to improve the model's coverage of different risk situations and enhance the model's generalization ability, so that the model can make relatively accurate risk predictions when facing new and unseen data.

[0072] Embodiment 2

[0073] Figure 2 This is a flow chart of another method for predicting financial risks of an enterprise provided in the second embodiment of the present invention. This embodiment is refined based on the above embodiment. Figure 2 As shown, the method specifically includes:

[0074] S210: Acquire the current operating data and current financial market environment data of the target enterprise to perform feature selection and / or feature extraction, and obtain the current original feature representation corresponding to the target enterprise.

[0075] S220: Input the current original feature representation into a pre-trained target encoder to obtain a current compact feature representation that matches the target enterprise.

[0076] S230, inputting the current compact feature representation into a pre-trained target ensemble learning framework.

[0077] S240. A target ensemble learning framework is used to perform weighted summation on individual financial risk prediction results output by multiple basic learners within the framework for the current compact feature representation, to obtain a target financial risk prediction result that matches the target enterprise, wherein different basic learners are constructed based on different reinforcement learning algorithms.

[0078] S250: If it is determined that the target financial risk prediction result contains a target financial risk, then obtain the current pending financial business application that matches the target enterprise.

[0079] S260: If the financial service application is associated with the target financial risk, issue a risk warning for the financial service application.

[0080] In the financial risk prediction process, through a series of processing (including obtaining the current operating data and current financial market environment data of the target enterprise, feature selection and / or feature extraction, using a pre-trained target encoder and a target ensemble learning framework for prediction, etc.), a target financial risk prediction result matching the target enterprise will eventually be obtained. When this prediction result indicates that the target enterprise has financial risks (that is, the prediction result contains the target financial risks), the next step is to obtain the financial business applications related to the target enterprise that are currently awaiting approval. These financial business applications may be submitted by the enterprise to banks or other financial institutions in order to conduct certain economic activities (such as loans, investments, financing, etc.).

[0081] After obtaining the target company's current pending financial business applications, it is necessary to determine whether these applications are associated with previously predicted financial risks. This association may be direct (such as loan applications for high-risk investment projects) or indirect (such as financing applications that may increase the company's financial pressure and thus increase financial risks). As a preventive measure, risk warning can take many forms, such as sending warning information to relevant personnel, marking risk levels in the system, and restricting or suspending the approval of certain financial businesses.

[0082] The technical solution of the embodiment of the present invention explains the follow-up tracking of financial risk prediction results by refining the overall solution. Specifically, risk correlation judgment is made for the target enterprise's currently awaiting approval financial business applications, and risk warning is issued when correlation occurs, thereby achieving a close integration of risk management with the enterprise's actual business processes, and being able to promptly remind the enterprise to pay attention to potential risks before the business is carried out, to avoid losses caused by blindly carrying out high-risk businesses, and to provide important risk reference basis for the enterprise's business decision-making; by early warning of risks, the enterprise can take proactive measures, such as adjusting business strategies, strengthening risk monitoring, increasing risk reserves, etc., to take pre-emptive precautions and control possible risks, which helps the enterprise reduce the probability and impact of risks, and enhance the enterprise's risk resistance in a complex financial environment.

[0083] Embodiment 3

[0084] Figure 3 This is a schematic diagram of the structure of a financial risk prediction device for an enterprise provided by Embodiment 3 of the present invention. Figure 3 As shown, the device comprises:

[0085] The original feature acquisition module 310 is used to acquire the current operation data of the target enterprise and the current financial market environment data to perform feature selection and / or feature extraction, and obtain the current original feature representation corresponding to the target enterprise;

[0086] A compact feature acquisition module 320 is used to input the current original feature representation into a pre-trained target encoder to obtain the current compact feature representation matching the target enterprise;

[0087] A feature input module 330, for inputting the current compact feature representation into a pre-trained target ensemble learning framework;

[0088] The prediction result acquisition module 340 is used to perform weighted summation on the individual financial risk prediction results output by multiple basic learners in the framework for the current compact feature representation through a target ensemble learning framework to obtain a target financial risk prediction result matching the target enterprise, wherein different basic learners are constructed based on different reinforcement learning algorithms.

[0089] The embodiment of the present invention starts with acquiring raw data, obtains raw feature representation through feature processing, further optimizes features through an encoder to obtain compact feature representation, then uses multiple basic learners in an integrated learning framework to make predictions, and finally weights and sums the prediction results of each learner to obtain the final risk prediction result. The entire process together constitutes a complete enterprise financial risk prediction method system, which fully integrates multi-source information so that the prediction model can learn different decision-making strategies to better adapt to market changes, and balances the advantages and limitations of different learners through the integration of multiple basic learners and weighted summation, thereby reducing the error of single model prediction, significantly improving the reliability of overall risk prediction, and reducing the risk of misjudgment.

[0090] Optionally, based on the above embodiments, it further includes an integrated learning framework construction unit, which is used to collect historical financial risk data of each enterprise, historical operation data associated with the historical financial risk data, and matched historical financial market environment data before inputting the current original feature representation into the pre-trained target encoder;

[0091] According to each historical operation data and the matching historical financial market environment data, the autoencoder model is trained, and the encoder in the trained autoencoder model is used as the target encoder;

[0092] According to the target encoder, the historical financial risk data of each enterprise, the historical operation data associated with the historical financial risk data, and the matching historical financial market environment data are processed to obtain a training sample set;

[0093] The training sample set is used to train each basic learner in the pre-built integrated learning framework, as well as the weighted weights of each basic learner when performing integrated prediction, to obtain the target integrated learning framework.

[0094] Optionally, based on the above embodiments, the integrated learning framework construction unit may further include:

[0095] A historical feature acquisition subunit is used to perform feature selection and / or feature extraction on each historical operation data and the matching historical financial market environment data to obtain multiple historical original feature representations;

[0096] A historical feature training subunit, used to input each historical original feature representation into a pre-built autoencoder model to train the autoencoder model;

[0097] The autoencoder model includes an encoder composed of a multi-layer neural network and a decoder symmetrical to the encoder network structure. The autoencoder model uses an adaptive moment estimation algorithm to minimize the reconstruction error.

[0098] Optionally, based on the above embodiments, the integrated learning framework construction unit may further include:

[0099] A historical compact feature acquisition subunit is used to input each historical operation data associated with each historical financial risk data and the matching historical financial market environment data into a target encoder to obtain a historical compact feature representation corresponding to each historical financial risk data;

[0100] The training sample set construction subunit is used to generate financial risk labels corresponding to each historical financial risk data, and combine each financial risk label with the matching historical compact feature representation to obtain a training sample set.

[0101] Optionally, on the basis of the above embodiments, it further includes an expanded sample collection unit, which is used to learn the training sample set through a self-supervised learning model before using the training sample set to train each basic learner in the pre-constructed integrated learning framework and the weighted weights of each basic learner when performing integrated prediction, wherein the self-supervised learning model is a generative adversarial network, and the generative adversarial network includes a generative model and a discriminative model;

[0102] After learning is completed, generating samples are collected from the generating model in the self-supervised learning model, and the generating samples are screened and / or corrected;

[0103] The generated samples that have completed the screening and / or correction process are used as expansion samples and added to the training sample set.

[0104] Optionally, on the basis of the above embodiments, it further includes a risk warning unit, which is used to obtain a current pending financial business application matching the target enterprise after obtaining the target financial risk prediction result matching the target enterprise, if it is determined that the target financial risk prediction result contains the target financial risk;

[0105] If the financial service application is associated with the target financial risk, a risk warning is issued for the financial service application.

[0106] A financial risk prediction device for an enterprise provided by an embodiment of the present invention can execute a financial risk prediction method for an enterprise provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

[0107] Embodiment 4

[0108] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0109] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0110] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0111] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The processor 11 executes the various methods and processes described above, such as a method for predicting financial risks of an enterprise.

[0112] That is: obtaining the current operating data of the target enterprise and the current financial market environment data for feature selection and / or feature extraction, and obtaining the current original feature representation corresponding to the target enterprise;

[0113] Input the current original feature representation into the pre-trained target encoder to obtain the current compact feature representation that matches the target enterprise;

[0114] Input the current compact feature representation into the pre-trained target ensemble learning framework;

[0115] Through the target ensemble learning framework, the individual financial risk prediction results output by multiple basic learners in the framework for the current compact feature representation are weighted summed to obtain the target financial risk prediction results matching the target enterprise, where different basic learners are constructed based on different reinforcement learning algorithms.

[0116] In some embodiments, a method for predicting financial risks of an enterprise may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for predicting financial risks of an enterprise described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute a method for predicting financial risks of an enterprise in any other appropriate manner (e.g., by means of firmware).

[0117] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0118] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0119] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0120] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0121] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0122] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0123] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0124] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting financial risks of an enterprise, characterized in that: include: Obtain the current operating data of the target enterprise and the current financial market environment data for feature selection and / or feature extraction to obtain the current original feature representation corresponding to the target enterprise; Input the current original feature representation into the pre-trained target encoder to obtain the current compact feature representation that matches the target enterprise; Input the current compact feature representation into the pre-trained target ensemble learning framework; Through the target ensemble learning framework, the individual financial risk prediction results output by multiple basic learners in the framework for the current compact feature representation are weighted summed to obtain the target financial risk prediction results matching the target enterprise, where different basic learners are constructed based on different reinforcement learning algorithms.

2. The method according to claim 1, characterized in that Before the current raw feature representation is input into the pre-trained target encoder, it also includes: Collect historical financial risk data of each enterprise, historical operating data associated with historical financial risk data, and matching historical financial market environment data; According to each historical operation data and the matching historical financial market environment data, the autoencoder model is trained, and the encoder in the trained autoencoder model is used as the target encoder; According to the target encoder, the historical financial risk data of each enterprise, the historical operation data associated with the historical financial risk data, and the matching historical financial market environment data are processed to obtain a training sample set; The training sample set is used to train each basic learner in the pre-built integrated learning framework, as well as the weighted weights of each basic learner when performing integrated prediction, to obtain the target integrated learning framework.

3. The method according to claim 2, characterized in that The autoencoder model is trained based on the historical operating data and the matching historical financial market environment data, including: Perform feature selection and / or feature extraction on each historical operation data and the matching historical financial market environment data to obtain multiple historical original feature representations; Input each historical original feature representation into the pre-built autoencoder model respectively, and train the autoencoder model; The autoencoder model includes an encoder composed of a multi-layer neural network and a decoder symmetrical to the encoder network structure. The autoencoder model uses an adaptive moment estimation algorithm to minimize the reconstruction error.

4. The method according to claim 2, characterized in that: According to the target encoder, the historical financial risk data of each enterprise, the historical operation data associated with the historical financial risk data, and the matching historical financial market environment data are processed to obtain a training sample set, including: Inputting each historical operation data respectively associated with each historical financial risk data and the matched historical financial market environment data into the target encoder respectively, and obtaining the historical compact feature representation corresponding to each historical financial risk data respectively; Generate financial risk labels corresponding to each historical financial risk data, and combine each financial risk label with the matching historical compact feature representation to obtain a training sample set.

5. The method according to claim 2, characterized in that: Before using the training sample set to train each base learner in the pre-built integrated learning framework and the weighted weights of each base learner when performing integrated prediction, it also includes: Learning the training sample set through a self-supervised learning model, wherein the self-supervised learning model is a generative adversarial network, and the generative adversarial network includes a generative model and a discriminative model; After learning is completed, generating samples are collected from the generating model in the self-supervised learning model, and the generating samples are screened and / or corrected; The generated samples that have completed the screening and / or correction process are used as expansion samples and added to the training sample set.

6. The method according to any one of claims 1 to 5, characterized in that: After obtaining the target financial risk prediction results that match the target enterprise, it also includes: If it is determined that the target financial risk prediction result contains the target financial risk, obtaining the currently pending financial business application matching the target enterprise; If the financial service application is associated with the target financial risk, a risk warning is issued for the financial service application.

7. A financial risk prediction device for an enterprise, characterized in that: include: The original feature acquisition module is used to obtain the current operating data of the target enterprise and the current financial market environment data for feature selection and / or feature extraction to obtain the current original feature representation corresponding to the target enterprise; A compact feature acquisition module, used to input the current original feature representation into a pre-trained target encoder to obtain the current compact feature representation matching the target enterprise; A feature input module, used to input the current compact feature representation into a pre-trained target ensemble learning framework; The prediction result acquisition module is used to perform weighted summation of individual financial risk prediction results output by multiple basic learners in the framework for the current compact feature representation through a target ensemble learning framework to obtain a target financial risk prediction result that matches the target enterprise, wherein different basic learners are constructed based on different reinforcement learning algorithms.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute a method for predicting financial risks of an enterprise as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement a financial risk prediction method for an enterprise according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the computer program implements a method for predicting financial risks of an enterprise according to any one of claims 1 to 6.

Images