A financial market risk prevention and control system and method based on a cross-language large model

Abstract

The present application provides a financial market risk prevention and control system and method based on a cross-language large model, the method comprising: clustering a financial market vector space to obtain a plurality of financial data cores; obtaining a language label corresponding to each financial market data vector in the financial market vector space, reconstructing the group behavior of the financial market vector space according to the group behavior prediction characteristics corresponding to each language label, to obtain a financial data reconstruction space; clustering the group behavior of the financial data reconstruction space to obtain a plurality of group behavior clustering cores, screening the plurality of group behavior clustering cores according to the spatial characteristics of all the group behavior clustering cores, to obtain a plurality of abnormal prediction centers; and screening the financial market data for risk data in real time according to the plurality of abnormal prediction centers, which can improve the accuracy of financial risk data screening by enhancing the feature contrast between the financial market data of users of different languages.

Description

Technical Field

[0001] The present application relates to the technical field of financial risk prevention and control, and more specifically, to a financial market risk prevention and control system and method based on a cross-language large model. Background Art

[0002] Financial market risk prevention and control refers to identifying, assessing, monitoring and managing various risks in the financial market to reduce their potential negative impact on the financial system, institutions and investors; the purpose of financial market risk prevention and control is to ensure the stability and healthy development of the financial system, protect the interests of investors and prevent the occurrence of financial crises. The financial market risk prevention and control system based on cross-language large models refers to the use of large-scale artificial intelligence models that can process multi-language data and cross-language information to enhance the risk management and prediction capabilities of the financial market.

[0003] The existing financial market risk prevention and control methods mainly establish a real-time monitoring system to track key market indicators (such as price, trading volume, volatility) and risk indicators (such as instantaneous financial risk, risk value), and set risk thresholds and trigger conditions. When the indicators exceed the preset range, early warning signals are automatically issued. This method is difficult to detect the in-depth behavioral information of financial market users of different languages, resulting in low feature contrast between financial market data of financial market users of different languages, resulting in low accuracy in financial risk data screening, and ultimately making the effect of financial market risk prevention and control poor. Summary of the invention

[0004] The present application provides a financial market risk prevention and control system and method based on a cross-language large model, which can improve the accuracy of financial risk data screening by enhancing the feature contrast between financial market data of users of different languages.

[0005] In a first aspect, the present application provides a financial market risk prevention and control method based on a cross-language large model. The method can be executed by a network device, or can also be executed by a chip configured in the network device, and the present application does not limit this.

[0006] Specifically, the method includes:

[0007] Obtain a financial market data group, divide the financial market data into multiple sub-language data groups based on a cross-language large model, perform group behavior prediction for different language groups based on different language labels corresponding to each sub-language data group, and obtain group behavior prediction features corresponding to each language label;

[0008] Performing digital vector conversion on the financial market data group to obtain a financial market vector space, and then performing data clustering on the financial market vector space to obtain a plurality of financial data cores;

[0009] Obtaining language labels corresponding to each financial market data vector in the financial market vector space, and reconstructing group behavior in the financial market vector space according to group behavior prediction features corresponding to each language label to obtain a financial data reconstruction space;

[0010] Performing group behavior clustering on the financial data reconstruction space to obtain a plurality of group behavior clustering cores, and obtaining a plurality of abnormal prediction centers after screening the group behavior clustering cores according to the spatial distribution characteristics of all group behavior clustering cores and all financial data cores;

[0011] Based on the multiple abnormal prediction centers, risk data screening is performed on financial market data in real time.

[0012] In conjunction with the first aspect, in certain implementations of the first aspect, dividing the financial market data into a plurality of sub-language data groups based on the cross-language macro model specifically includes:

[0013] Obtain a preset language classification test data set, which includes text samples in different languages ​​and their corresponding labels;

[0014] A support vector machine is used as a large cross-language model, and the support vector machine is trained by using the language classification test data set;

[0015] Use the trained cross-language large model to perform language recognition on the text content in the financial market data, and divide the financial market data into multiple sub-language data groups based on the recognized language labels.

[0016] In combination with the first aspect, in certain implementations of the first aspect, group behavior prediction of different language groups is performed based on different language labels corresponding to each sub-language data group, and group behavior prediction features corresponding to each language label are obtained, specifically including:

[0017] Extract group behavior characteristics from each sub-language data group, including price fluctuations, trading volume changes, sentiment analysis results, etc.

[0018] A long short-term memory neural network training model is selected to train each sub-language data group to obtain the user behavior prediction features corresponding to each sub-language data group;

[0019] The language labels corresponding to each sub-language data group are obtained, and all user behavior prediction features are digitized according to the language labels corresponding to each sub-language data group to obtain group behavior prediction features corresponding to each language label.

[0020] In combination with the first aspect, in certain implementations of the first aspect, performing digital vector conversion on the financial market data group to obtain a financial market vector space specifically includes: using a bag-of-words model to convert each financial market data in the financial market data group into a corresponding numerical vector to obtain each financial market data vector; and combining all financial market data vectors to obtain the financial market vector space.

[0021] In combination with the first aspect, in certain implementations of the first aspect, after screening the group behavior clustering cores by using the spatial distribution characteristics of all group behavior clustering cores and all financial data cores, the multiple abnormal prediction centers obtained specifically include:

[0022] Acquire a first group behavior clustering kernel, and determine a nearest financial data kernel of the first group behavior clustering kernel;

[0023] Obtaining other group behavior clustering cores except the first group behavior clustering core, repeating the above steps, and obtaining the nearest financial data cores corresponding to each group behavior clustering core;

[0024] Determine whether there is a financial data core that is the nearest neighbor financial data core corresponding to multiple group behavior clustering cores, screen out the group behavior data cores corresponding to all financial data cores of this type, and obtain multiple market risk high-dimensional centers;

[0025] According to the group behavior characteristic values ​​corresponding to each high-dimensional market risk center, data dimensionality reduction is performed on the high-dimensional market risk center to obtain multiple abnormal prediction centers.

[0026] In conjunction with the first aspect, in certain implementations of the first aspect, performing risk data screening on financial market data in real time based on multiple abnormality prediction centers specifically includes:

[0027] Acquire real-time financial market data, and perform digital vector conversion on the real-time financial market data to obtain a real-time financial market data vector;

[0028] Determining data risk according to characteristic deviations corresponding to the real-time data vector of the financial market and each abnormal prediction center;

[0029] According to the data risk level, it is determined whether the real-time financial market data is risk data, and the risk data is screened out.

[0030] In combination with the first aspect, in certain implementations of the first aspect, before dividing the financial market data into a plurality of sub-language data groups based on a cross-language big model, the method further includes: performing data preprocessing on the financial market data group.

[0031] In a second aspect, the present application provides a financial market risk prevention and control system based on a cross-language large model, including a risk prevention and control unit, wherein the risk prevention and control unit includes:

[0032] An acquisition module is used to acquire a financial market data group, divide the financial market data into multiple sub-language data groups based on a cross-language large model, and perform group behavior prediction for different language groups based on different language labels corresponding to each sub-language data group to obtain group behavior prediction features corresponding to each language label;

[0033] A processing module, configured to perform digital vector conversion on the financial market data group to obtain a financial market vector space, and then perform data clustering on the financial market vector space to obtain a plurality of financial data cores;

[0034] The processing module is further used to obtain language labels corresponding to each financial market data vector in the financial market vector space, and reconstruct group behavior in the financial market vector space according to group behavior prediction features corresponding to each language label to obtain a financial data reconstruction space;

[0035] The processing module is further used to perform group behavior clustering on the financial data reconstruction space to obtain a plurality of group behavior clustering cores, and to obtain a plurality of abnormal prediction centers after screening the group behavior clustering cores according to the spatial distribution characteristics of all group behavior clustering cores and all financial data cores;

[0036] An execution module is used to perform risk data screening on financial market data in real time according to the multiple abnormal prediction centers.

[0037] In a third aspect, the present application provides a computer terminal device, comprising a memory and a processor, wherein the memory stores code, and the processor is configured to obtain the code and execute the above-mentioned financial market risk prevention and control method based on a cross-language large model.

[0038] In a fourth aspect, the present application provides a computer-readable storage medium, which stores at least one computer program, and the computer program is loaded and executed by a processor to implement the operations performed by the above-mentioned financial market risk prevention and control method based on a cross-language large model.

[0039] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects:

[0040] In a financial market risk prevention and control system and method based on a cross-language large model provided by the present application, a financial market data group is first obtained, and the financial market data is divided into multiple sub-language data groups based on the cross-language large model, and group behavior prediction of different language groups is performed according to different language labels corresponding to each sub-language data group to obtain group behavior prediction features corresponding to each language label; the financial market data group is converted into a digital vector to obtain a financial market vector space, and the financial market vector space is clustered to obtain multiple financial data cores; the language labels corresponding to each financial market data vector in the financial market vector space are obtained, and the group behavior of the financial market vector space is reconstructed according to the group behavior prediction features corresponding to each language label to obtain a financial data reconstruction space; the financial data reconstruction space is clustered to obtain multiple group behavior clustering cores, and the multiple group behavior clustering cores are screened according to the spatial features of all group behavior clustering cores and all financial data cores to obtain multiple abnormal prediction centers; and the financial market data is screened for risk data in real time according to the multiple abnormal prediction centers.

[0041] In summary, based on the significant differences in group behavior of users of different languages ​​in the financial market, this application captures and analyzes group behavior through a cross-language model, obtains group behavior prediction features corresponding to each language label, and thus makes more accurate behavior predictions, and then clusters the financial market vector space to obtain multiple financial data cores, and reconstructs group behavior in the financial market vector space according to the group behavior prediction features corresponding to each language label, thereby constructing a multidimensional financial data reconstruction space, which integrates the prediction results of group behavior in each language, enhances the feature contrast between financial market data of users of different languages, and enables the system to identify potential market fluctuations and risk points, clusters group behavior in the reconstruction space, and screens the abnormal prediction center in combination with the features of the financial data core, which can more prominently display abnormal behavior and potential risk areas in the market. This application enhances the contrast of user data of different languages ​​by high-dimensional features based on multilingual user behavior features, facilitates risk identification, and finally can screen out risk data in financial market data in real time according to the determined abnormal prediction center, thereby improving the accuracy of financial risk data screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is an exemplary flow chart of a financial market risk prevention and control method based on a cross-language large model according to some embodiments of the present application;

[0043] Figure 2 is an exemplary flow chart for determining group behavior prediction features corresponding to each language label in some embodiments of the present application;

[0044] Figure 3 is a schematic diagram of exemplary hardware and / or software of a risk prevention and control unit according to some embodiments of the present application;

[0045] Figure 4 It is a structural schematic diagram of a computer terminal device for implementing a financial market risk prevention and control method based on a cross-language large model according to some embodiments of the present application. DETAILED DESCRIPTION

[0046] The present application obtains a financial market data group, divides the financial market data into multiple sub-language data groups based on a cross-language large model, predicts the group behavior of different language groups according to different language labels corresponding to each sub-language data group, and obtains the group behavior prediction features corresponding to each language label; performs digital vector conversion on the financial market data group to obtain a financial market vector space, clusters the financial market vector space to obtain multiple financial data cores; obtains the language labels corresponding to each financial market data vector in the financial market vector space, reconstructs the group behavior of the financial market vector space according to the group behavior prediction features corresponding to each language label, and obtains a financial data reconstruction space; clusters the group behavior of the financial data reconstruction space to obtain multiple group behavior clustering cores, and screens the multiple group behavior clustering cores according to the spatial features of all group behavior clustering cores and all financial data cores to obtain multiple abnormal prediction centers; and performs risk data screening on the financial market data in real time according to the multiple abnormal prediction centers, which can improve the accuracy of financial risk data screening by enhancing the feature contrast between the financial market data of users of different languages.

[0047] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 , which is an exemplary flow chart of a financial market risk prevention and control method based on a cross-language big model according to some embodiments of the present application. The financial market risk prevention and control method 100 based on a cross-language big model mainly includes the following steps:

[0048] In step S101, a financial market data group is obtained, and the financial market data is divided into multiple sub-language data groups based on a cross-language large model. Group behavior prediction of different language groups is performed according to different language labels corresponding to each sub-language data group, and group behavior prediction features corresponding to each language label are obtained.

[0049] Optionally, in some embodiments, an online financial forum can be used as the data source of the financial market data group. In specific implementation, a web crawler tool can be used to collect the required financial market data, wherein the data type of the financial market data group may include stock prices, trading volume, news, social media comments, etc.

[0050] Optionally, in some embodiments, before dividing the financial market data into multiple sub-language data groups based on the cross-language big model, the method may also include: performing data preprocessing on the financial market data group. For example, natural language processing (NLP) operations such as word segmentation, stop word removal, and stem extraction may be used to achieve data preprocessing of the financial market data group.

[0051] Preferably, in some embodiments, dividing the financial market data into multiple sub-language data groups based on the cross-language macro model can be achieved by using the following steps:

[0052] Obtain a preset language classification test data set, which includes text samples in different languages ​​and their corresponding labels;

[0053] A support vector machine is used as a large cross-language model, and the support vector machine is trained by using the language classification test data set;

[0054] Use the trained cross-language large model to perform language recognition on the text content in the financial market data, and divide the financial market data into multiple sub-language data groups (such as English market data group, Chinese market data group, etc.) based on the recognized language labels.

[0055] Optionally, in some embodiments, reference Figure 2 As shown in FIG. 1 , this figure is an exemplary flow chart for determining group behavior prediction features corresponding to each language label in some embodiments of the present application. Group behavior prediction of different language groups is performed based on different language labels corresponding to each sub-language data group. The group behavior prediction features corresponding to each language label can be obtained by using the following steps:

[0056] In step S1011, group behavior features are extracted from each sub-language data group. In specific implementation, the group behavior features include price fluctuations, transaction volume changes, sentiment analysis results, etc. of each sub-language data group;

[0057] In step S1012, a long short-term memory neural network (LSTM) training model is selected to train the group behavior characteristics corresponding to each sub-language data group to obtain user behavior prediction characteristics corresponding to each sub-language data group;

[0058] In step S1013, the language labels corresponding to the sub-language data groups are obtained, and all user behavior prediction features are digitized according to the language labels corresponding to the sub-language data groups to obtain group behavior prediction features corresponding to the language labels.

[0059] It should be noted that the LSTM training model is a special recursive neural network (RNN), which is commonly used in the prior art to process and predict time series data. The present application uses the historical data of each sub-language data group to train the LSTM model to predict future group behavior. In specific implementation, the extracted behavior features can be input into the LSTM model, and the model optimizes the model parameters through back propagation and gradient descent so that it can accurately predict group behavior. In some embodiments, the digitization of all user behavior prediction features can be achieved through a preset feature mapping table to obtain the group behavior prediction features corresponding to each language label. In some other embodiments, the distance value between each user behavior prediction feature and the preset standard user behavior prediction feature in the feature space can also be used as the group behavior prediction feature.

[0060] In step S102, digital vector conversion is performed on the financial market data group to obtain a financial market vector space, and data clustering is performed on the financial market vector space to obtain multiple financial data cores.

[0061] It should be noted that the financial market vector space is a vector set including multiple financial market data vectors, and the financial market data vector is a numerical vector converted from financial market data, wherein each financial market data vector has its unique corresponding financial market data. Optionally, in some embodiments, the financial market data group is converted into a digital vector to obtain a financial market vector space, which can be achieved by the following steps:

[0062] The bag-of-words model is used to convert each financial market data in the financial market data group into a corresponding numerical vector to obtain each financial market data vector; all financial market data vectors are combined to obtain the financial market vector space.

[0063] Optionally, in some embodiments, the K-means clustering method can be used to cluster the financial market vector space to obtain multiple financial data cores; in specific implementation, the K-means clustering method can be used to cluster each dimension of each financial market data vector in the financial market vector space to obtain multiple clustering centers as the financial data cores. For example, after obtaining K clustering results in each dimension, the values ​​corresponding to each clustering result are respectively formed into a clustering center according to the time series to obtain K clustering centers. In specific implementation, the K value can be calibrated as a constant according to the number of data in the financial market data group, which will not be repeated here.

[0064] In step S103, the language label corresponding to each financial market data vector in the financial market vector space is obtained, and the group behavior of the financial market vector space is reconstructed according to the group behavior prediction characteristics corresponding to each language label to obtain a financial data reconstruction space.

[0065] Optionally, in some embodiments, the language labels corresponding to the financial market data vectors may be obtained according to the mapping relationship between the financial market data in the financial market data group and the financial market data vectors and according to the classification results of the sub-language data groups.

[0066] It should be noted that the financial data reconstruction space is a vector space containing multiple financial market data prediction vectors, and the financial market data prediction vector is a data vector used to reflect the behavioral tendencies of users of different languages ​​after the financial market data vector is expanded by vector dimensionality according to the group behavior prediction features corresponding to each language label. Optionally, in some embodiments, the financial market vector space is reconstructed by group behavior according to the group behavior prediction features corresponding to each language label to obtain the financial data reconstruction space, which can be achieved by the following steps:

[0067] According to the group behavior prediction features corresponding to each language label, each financial market data vector in the financial market vector space is vector-dimensionalized to obtain multiple financial market data prediction vectors, and the financial data reconstruction space is composed of all the financial market data prediction vectors.

[0068] Preferably, in some embodiments, according to the group behavior prediction features corresponding to each language label, the vector dimensionality of each financial market data vector in the financial market vector space can be increased by the following steps: adding an additional dimension to the financial market data vector, embedding the language label and the group behavior prediction features into the original financial data vector, so as to obtain a higher dimensional financial market data prediction vector.

[0069] In step S104, group behavior clustering is performed on the financial data reconstruction space to obtain a plurality of group behavior clustering cores, and the group behavior clustering cores are screened by spatial distribution characteristics of all group behavior clustering cores and all financial data cores to obtain a plurality of abnormal prediction centers.

[0070] Optionally, in some embodiments, a K-means clustering method may be used to perform group behavior clustering on the financial data reconstruction space to obtain a plurality of group behavior clustering kernels.

[0071] Preferably, in some embodiments, performing group behavior clustering on the financial data reconstruction space to obtain multiple group behavior clustering cores can be achieved by using the following steps:

[0072] Get the value of financial data core;

[0073] Initializing the financial data reconstruction space according to the number of financial data cores to obtain a plurality of initial clustering cores;

[0074] For each financial market data prediction vector in the financial data reconstruction space, assign it to an initial clustering core of the nearest neighbor distance, thereby obtaining a plurality of initial clustering clusters;

[0075] For each initial cluster, calculate the mean of all samples in the cluster and use the mean as the updated initial cluster;

[0076] Repeat the above steps until the updated initial clustering cluster remains unchanged or the change amplitude is lower than the preset change threshold, and use each initial clustering cluster as multiple group behavior clustering cores.

[0077] It should be noted that, in the present application, the method of screening the group behavior clustering cores by the spatial distribution characteristics of all group behavior clustering cores and all financial data cores can filter the repeated features in the group behavior prediction according to the characteristic deviations of the group behavior clustering cores and the financial data cores, thereby extracting effective group behavior clustering cores as abnormal prediction centers and performing financial market risk management, and ultimately increasing the accuracy of financial market risk prediction. Preferably, in some embodiments, after screening the group behavior clustering cores according to the spatial distribution characteristics of all group behavior clustering cores and all financial data cores, multiple abnormal prediction centers can be obtained, which can be achieved by the following steps:

[0078] Acquire a first group behavior clustering kernel, and determine a nearest financial data kernel of the first group behavior clustering kernel;

[0079] Obtaining other group behavior clustering cores except the first group behavior clustering core, repeating the above steps, and obtaining the nearest financial data cores corresponding to each group behavior clustering core;

[0080] Determine whether there is a financial data core that is the nearest neighbor financial data core corresponding to multiple group behavior clustering cores, screen out the group behavior data cores corresponding to all financial data cores of this type, and obtain multiple market risk high-dimensional centers;

[0081] According to the group behavior characteristic values ​​corresponding to each high-dimensional market risk center, data dimensionality reduction is performed on the high-dimensional market risk center to obtain multiple abnormal prediction centers.

[0082] It should be noted that the abnormal prediction center is an abnormal data comparison center when the financial market data is screened for abnormal risk data. Preferably, in some embodiments, the market risk high-dimensional center is subjected to data dimensionality reduction according to the group behavior characteristic values ​​corresponding to each market risk high-dimensional center, and the acquisition of multiple abnormal prediction centers can be achieved by the following steps:

[0083] Obtain the maximum dimensional element of each market risk high-dimensional center as the corresponding group behavior characteristic value;

[0084] Obtain the prediction center deviation corresponding to each market risk high-dimensional center according to the group behavior characteristic value corresponding to each market risk high-dimensional center. In specific implementation, the prediction center deviation of the market risk high-dimensional center = |mean of group behavior characteristic value - corresponding group behavior characteristic value| / mean of group behavior characteristic value;

[0085] Data dimensionality reduction is performed on the high-dimensional center of market risk to obtain multiple abnormal prediction centers, and a corresponding prediction center deviation is assigned to each abnormal prediction center.

[0086] It should be noted that by performing data dimensionality reduction on the high-dimensional market risk centers according to the group behavior characteristic values ​​corresponding to each high-dimensional market risk center, different abnormal prediction centers can be assigned different prediction center deviations based on the degree of deviation between the clustering results and the mean of the clustering results, thereby distinguishing between important and secondary abnormal prediction centers and increasing the accuracy of financial market risk management and control.

[0087] Optionally, in some embodiments, the anomaly prediction center can be obtained by performing data restoration on the market risk high-dimensional center. For example, the maximum dimensional element of the market risk high-dimensional center is deleted to obtain a new reduced-dimensional data vector as the anomaly prediction center.

[0088] In step S105, risk data screening is performed on financial market data in real time based on multiple abnormal prediction centers.

[0089] Optionally, in some embodiments, based on multiple abnormal prediction centers, real-time risk data screening of financial market data can be implemented by the following steps:

[0090] Acquire real-time financial market data, and perform digital vector conversion on the real-time financial market data to obtain a real-time financial market data vector;

[0091] Determining data risk according to characteristic deviations corresponding to the real-time data vector of the financial market and each abnormal prediction center;

[0092] According to the data risk level, it is determined whether the real-time financial market data is risk data, and the risk data is screened out.

[0093] Optionally, in some embodiments, according to the characteristic deviations corresponding to the real-time financial market data vector and each abnormal prediction center, determining the data risk can be achieved by the following steps:

[0094] Obtain the prediction center deviations corresponding to the respective abnormal prediction centers, and obtain the characteristic deviations corresponding to the real-time data vector of the financial market and the respective abnormal prediction centers;

[0095] According to the prediction center deviations corresponding to the abnormal prediction centers, the real-time data vector of the financial market and the characteristic deviations corresponding to the abnormal prediction centers are weightedly fused to obtain the data risk degree.

[0096] It should be noted that when the data risk of the real-time financial market data is relatively high, it indicates that there is a large deviation between the real-time financial market data and the behavioral habits of financial market users in various languages. The real-time financial market data may be artificially modified or computer-generated. It is necessary to delete the real-time financial market data in advance to avoid the impact of abnormal financial market data and achieve financial market risk prevention and control. Optionally, in some embodiments, judging whether the real-time financial market data is risk data according to the data risk includes:

[0097] Determine whether the data risk is higher than a preset risk threshold. If the data risk is higher than the preset risk threshold, determine that the financial market real-time data is risk data; otherwise, determine that the financial market real-time data is not risk data.

[0098] In addition, in another aspect of the present application, in some embodiments, the present application provides a financial market risk prevention and control system based on a cross-language large model, the system comprising a risk prevention and control unit, reference Figure 3 , which is a schematic diagram of exemplary hardware and / or software of a risk prevention and control unit according to some embodiments of the present application, the risk prevention and control unit 200 includes: an acquisition module 201, a processing module 202 and an execution module 203, which are described as follows:

[0099] Acquisition module 201, in some specific embodiments of the present application, acquisition module 201 is mainly used to acquire a financial market data group, divide the financial market data into multiple sub-language data groups based on a cross-language macro model, perform group behavior prediction for different language groups based on different language labels corresponding to each sub-language data group, and obtain group behavior prediction features corresponding to each language label;

[0100] In some specific embodiments of the present application, the processing module 202 is used to perform digital vector conversion on the financial market data group to obtain a financial market vector space, and then perform data clustering on the financial market vector space to obtain multiple financial data cores;

[0101] The processing module 202 is further used to obtain language labels corresponding to each financial market data vector in the financial market vector space, and reconstruct group behavior in the financial market vector space according to group behavior prediction features corresponding to each language label to obtain a financial data reconstruction space;

[0102] The processing module 202 is further used to perform group behavior clustering on the financial data reconstruction space to obtain a plurality of group behavior clustering cores, and to obtain a plurality of abnormal prediction centers after screening the group behavior clustering cores according to the spatial distribution characteristics of all group behavior clustering cores and all financial data cores;

[0103] Execution module 203, in some specific embodiments of the present application, execution module 203 is mainly used to perform risk data screening on financial market data in real time according to the multiple abnormal prediction centers.

[0104] The above describes in detail an example of a financial market risk prevention and control system and method based on a cross-language large model provided in an embodiment of the present application. It can be understood that in order to realize the above functions, the corresponding device includes a hardware structure and / or software module corresponding to executing each function.

[0105] Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed in this document, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function in the application is executed in hardware or in a computer software-driven hardware manner depends on the specific application and design constraints of the technical solution. Therefore, professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0106] In addition, the present application also provides a computer terminal device, which includes a memory and a processor, the memory stores code, and the processor is configured to obtain the code and execute the above-mentioned financial market risk prevention and control method based on a cross-language large model.

[0107] In some embodiments, reference Figure 4 , which is a schematic diagram of the structure of a computer terminal device for implementing a financial market risk prevention and control method based on a cross-language large model according to some embodiments of the present application. A financial market risk prevention and control method based on a cross-language large model in the above embodiment can be achieved by Figure 4 The computer terminal device 300 shown in the figure is implemented, and the computer terminal device 300 includes at least one communication bus 301, a communication interface 302, a processor 303 and a memory 304.

[0108] Processor 303 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more processors for controlling the execution of a financial market risk prevention and control method based on a cross-language large model in the present application.

[0109] The communication bus 301 may include a path for transmitting information between the above-mentioned components.

[0110] The memory 304 may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compressed optical disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory 304 may exist independently and be connected to the processor 303 via the communication bus 301. The memory 304 may also be integrated with the processor 303.

[0111] The memory 304 is used to store the program code for executing the solution of the present application, and the execution is controlled by the processor 303. The processor 303 is used to execute the program code stored in the memory 304. The program code may include one or more software modules. The determination of the group behavior prediction feature in the above embodiment can be implemented by the processor 303 and one or more software modules in the program code in the memory 304.

[0112] The communication interface 302 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0113] Optionally, the computer terminal device 300 may further include a power supply 305 for providing power to various devices or circuits in the real-time computer terminal device.

[0114] In a specific implementation, as an embodiment, a computer terminal device may include multiple processors, each of which may be a single-CPU processor or a multi-CPU processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0115] The above-mentioned computer terminal device may be a general-purpose computer terminal device or a dedicated computer terminal device. In a specific implementation, the computer terminal device may be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device or an embedded device. The embodiment of the present application does not limit the type of computer terminal device.

[0116] In addition, in other aspects of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores at least one computer program, and the computer program is loaded and executed by a processor to implement the operations performed by the above-mentioned financial market risk prevention and control method based on a cross-language large model.

[0117] In summary, in a financial market risk prevention and control system and method based on a cross-language big model disclosed in an embodiment of the present application, firstly, by acquiring a financial market data group, the financial market data is divided into multiple sub-language data groups based on a cross-language big model, and group behavior prediction of different language groups is performed according to different language labels corresponding to each sub-language data group, and group behavior prediction features corresponding to each language label are obtained; the financial market data group is converted into a digital vector to obtain a financial market vector space, and the financial market vector space is clustered to obtain multiple financial data cores; the language labels corresponding to each financial market data vector in the financial market vector space are acquired, and group behavior is reconstructed on the financial market vector space according to the group behavior prediction features corresponding to each language label to obtain a financial data reconstruction space; group behavior is clustered on the financial data reconstruction space to obtain multiple group behavior clustering cores, and multiple group behavior clustering cores are screened according to the spatial features of all group behavior clustering cores to obtain multiple abnormal prediction centers; based on the multiple abnormal prediction centers, risk data screening is performed on the financial market data in real time, which can improve the accuracy of financial risk data screening by enhancing the feature contrast between the financial market data of users of different languages.

[0118] The above is only an embodiment of the present application, and the common knowledge such as the specific technical scheme or characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, without departing from the technical scheme of the present application, several modifications and improvements can be made, which should also be regarded as the scope of protection of the present application, and these will not affect the effect of the implementation of the present application and the practicality of the patent.

[0119] The scope of protection claimed by this application shall be based on the content of its claims. The specific implementation methods and other records in the specification can be used to interpret the content of the claims. Obviously, those skilled in the art can make various changes and modifications to this application without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include these modifications and variations.

Claims

1. A financial market risk prevention and control method based on a cross-language large model, characterized in that: include: Obtain a financial market data group, divide the financial market data into multiple sub-language data groups based on a cross-language large model, perform group behavior prediction for different language groups based on different language labels corresponding to each sub-language data group, and obtain group behavior prediction features corresponding to each language label; Performing digital vector conversion on the financial market data group to obtain a financial market vector space, and then performing data clustering on the financial market vector space to obtain a plurality of financial data cores; Obtaining language labels corresponding to each financial market data vector in the financial market vector space, and reconstructing group behavior in the financial market vector space according to group behavior prediction features corresponding to each language label to obtain a financial data reconstruction space; Performing group behavior clustering on the financial data reconstruction space to obtain a plurality of group behavior clustering cores, and obtaining a plurality of abnormal prediction centers after screening the group behavior clustering cores according to the spatial distribution characteristics of all group behavior clustering cores and all financial data cores; Based on the multiple abnormal prediction centers, risk data screening is performed on financial market data in real time; Among them, online financial forums are used as the data source of the financial market data group, and web crawling tools are used to collect financial market data, and the financial market data group is social media comment data; Among them, the group behavior prediction of different language groups is performed according to the different language labels corresponding to each sub-language data group, and the group behavior prediction features corresponding to each language label are obtained, which specifically include: Extracting group behavior features from each sub-language data group, wherein the group behavior features include changes in transaction volume of each sub-language data group; The long short-term memory neural network training model is used to train the group behavior characteristics of each sub-language data group to obtain the user behavior prediction characteristics corresponding to each sub-language data group; Obtaining language labels corresponding to each sub-language data group, digitizing all user behavior prediction features according to the language labels corresponding to each sub-language data group, and obtaining group behavior prediction features corresponding to each language label, wherein the digitization of all user behavior prediction features is achieved through a preset feature mapping table, and the group behavior prediction features corresponding to each language label are obtained; Among them, after screening the group behavior clustering cores through the spatial distribution characteristics of all group behavior clustering cores and all financial data cores, multiple abnormal prediction centers are obtained, including: Acquire a first group behavior clustering kernel, and determine a nearest financial data kernel of the first group behavior clustering kernel; Obtaining other group behavior clustering cores except the first group behavior clustering core, repeating the above steps, and obtaining the nearest financial data cores corresponding to each group behavior clustering core; Determine whether there is a financial data core that is the nearest neighbor financial data core corresponding to multiple group behavior clustering cores, screen out the group behavior data cores corresponding to all financial data cores of this type, and obtain multiple market risk high-dimensional centers; According to the group behavior characteristic values ​​corresponding to each high-dimensional market risk center, data dimensionality reduction is performed on the high-dimensional market risk center to obtain multiple abnormal prediction centers; Among them, based on the multiple abnormal prediction centers, real-time risk data screening of financial market data specifically includes: Acquire real-time financial market data, and perform digital vector conversion on the real-time financial market data to obtain a real-time financial market data vector; Determining data risk according to characteristic deviations corresponding to the real-time data vector of the financial market and each abnormal prediction center; According to the data risk level, it is determined whether the real-time financial market data is risk data, and the risk data is screened out.

2. The method according to claim 1, characterized in that Based on the cross-language big model, the financial market data is divided into multiple sub-language data groups, including: Obtain a preset language classification test data set, where the language classification test data set includes text samples in different languages ​​and their corresponding labels; A support vector machine is used as a large cross-language model, and the support vector machine is trained by using the language classification test data set; Use the trained cross-language large model to perform language recognition on the text content in the financial market data, and divide the financial market data into multiple sub-language data groups based on the recognized language labels.

3. The method according to claim 1, characterized in that Converting the financial market data group into digital vectors to obtain the financial market vector space specifically includes: using a bag-of-words model to convert each financial market data in the financial market data group into a corresponding digital vector to obtain each financial market data vector; and combining all the financial market data vectors to obtain the financial market vector space.

4. The method according to claim 1, characterized in that Before dividing the financial market data into a plurality of sub-language data groups based on a cross-language large model, the method further includes: performing data preprocessing on the financial market data groups.

5. A financial market risk prevention and control system based on a cross-language large model, which is controlled by the method described in any one of claims 1 to 4, and the system includes a risk prevention and control unit, characterized in that: The risk prevention and control unit includes: An acquisition module is used to acquire a financial market data group, divide the financial market data into multiple sub-language data groups based on a cross-language large model, and perform group behavior prediction for different language groups based on different language labels corresponding to each sub-language data group to obtain group behavior prediction features corresponding to each language label; A processing module, configured to perform digital vector conversion on the financial market data group to obtain a financial market vector space, and then perform data clustering on the financial market vector space to obtain a plurality of financial data cores; The processing module is further used to obtain language labels corresponding to each financial market data vector in the financial market vector space, and reconstruct group behavior in the financial market vector space according to group behavior prediction features corresponding to each language label to obtain a financial data reconstruction space; The processing module is further used to perform group behavior clustering on the financial data reconstruction space to obtain a plurality of group behavior clustering cores, and to obtain a plurality of abnormal prediction centers after screening the group behavior clustering cores according to the spatial distribution characteristics of all group behavior clustering cores and all financial data cores; An execution module is used to perform risk data screening on financial market data in real time according to the multiple abnormal prediction centers.

6. A computer terminal device, characterized in that: The computer terminal device includes a memory and a processor, the memory stores codes, and the processor is configured to obtain the codes and execute a financial market risk prevention and control method based on a cross-language large model as described in any one of claims 1 to 4.

7. A computer-readable storage medium storing at least one computer program, characterized in that: The computer program is loaded and executed by a processor to implement the operations performed by a financial market risk prevention and control method based on a cross-language large model as described in any one of claims 1 to 4.

Images