Bank-to-bank bond market abnormal fluctuation risk monitoring system and method

Abstract

The invention relates to the technical field of financial science and technology and risk management, and discloses an inter-bank bond market abnormal fluctuation risk monitoring system and method. The system comprises a multi-source data fusion module, a dynamic risk factor engine module, an intelligent early warning model module, a pressure conduction simulation module, a visual decision support module and a supervision co-processing module. Multi-source heterogeneous data are collected in real time through a distributed crawler framework, dynamic analysis and anomaly detection of risk factors are realized in combination with an adaptive weight algorithm and a space-time diagram neural network, and a risk conduction path under an extreme scene is simulated by adopting a multi-agent modeling technology. And risk early warning and co-processing are realized by means of three-dimensional visualization and intelligent contract technologies. According to the invention, the problems of difficulty in data integration, early warning lagging, unclear conduction path and the like in traditional bond market risk monitoring are solved, and full-dimension data fusion, real-time dynamic monitoring, accurate risk early warning and intelligent co-processing are realized.

Description

Technical Field

[0001] The present invention relates to the fields of financial technology and risk management technology, and in particular to a system and method for monitoring abnormal fluctuation risks in the interbank bond market. Background Art

[0002] In recent years, the interbank bond market has continued to expand, with increasingly complex trading products and participating institutions, significantly increasing the risk of market volatility. Traditional risk monitoring methods, relying primarily on single-dimensional trading data analysis, struggle to comprehensively capture early signals of abnormal market fluctuations. With the prevalence of new trading methods such as high-frequency trading and algorithmic trading, market fluctuations have exhibited new characteristics such as nonlinearity and suddenness, posing significant challenges to existing monitoring systems.

[0003] Existing technologies for bond market risk monitoring primarily face the following challenges: limited data collection, focusing solely on transaction data and failing to integrate heterogeneous data from multiple sources, such as public opinion and the macroeconomy, resulting in a narrow perspective for risk identification; static weighting in risk factor models prevents them from dynamically responding to changes in market microstructure, leading to delayed risk assessment results; and early warning mechanisms rely on traditional quantitative models, which struggle to capture risk transmission pathways in complex network environments, resulting in insufficiently accurate early warning signals. More significantly, existing systems lack effective stress testing tools to simulate risk contagion processes under extreme scenarios. Visualization methods are limited, making it difficult to provide intuitive support for regulatory decision-making. Furthermore, the level of automation in risk management processes is low, and response efficiency fails to meet the demands of real-time supervision. Current technologies also suffer from insufficient data processing capabilities, resulting in significant system response delays when faced with massive amounts of high-frequency trading data. Poor data flow between functional modules creates information silos, hindering the integrity and coordination of risk monitoring. These shortcomings severely hinder the timeliness and effectiveness of bond market risk monitoring. Summary of the Invention

[0004] The present invention provides a method and system for monitoring the risk of abnormal fluctuations in the interbank bond market, the main purpose of which is to solve the problems of insufficient data integration, delayed early warning and unclear risk transmission path in traditional risk monitoring methods in the interbank bond market.

[0005] To achieve the above-mentioned objectives, the present invention provides a system for monitoring the risk of abnormal fluctuations in the interbank bond market. The system comprises a multi-source data fusion module, a dynamic risk factor engine module, an intelligent early warning model module, a pressure transmission simulation module, a visual decision support module, and a supervisory collaborative disposal module, wherein: The multi-source data fusion module is used to collect and clean the transaction flow, macroeconomic indicators and public opinion texts of the interbank bond market in real time based on a distributed crawler framework to obtain a standardized spatiotemporal series data set; The dynamic risk factor engine module is used to perform an adaptive weighting algorithm analysis on market microstructure characteristics based on a standardized spatiotemporal series data set to obtain a time-varying risk parameter matrix; The intelligent early warning model module is used to perform spatiotemporal graph neural network analysis on standardized risk indicators based on a time-varying risk parameter matrix to obtain a set of early warning signals with probability evaluation; The pressure transmission simulation module is used to perform multi-agent modeling and deduction of the balance sheet of the institution under extreme scenarios based on the early warning signal set to obtain the risk contagion heat map; The visualization decision support module is used to convert market risk parameters into a three-dimensional rendering engine based on the risk contagion heat map to obtain an interactive supervision view; The regulatory collaborative disposal module is used to trigger smart contract rules for confirmed abnormal events based on the interactive regulatory view to obtain the execution results of the hierarchical response strategy.

[0006] Optionally, when the multi-source data fusion module performs real-time collection of transaction flows, macroeconomic indicators, and public opinion texts of the interbank bond market based on a distributed crawler framework, the multi-source data fusion module includes: Relying on a distributed crawler cluster to crawl text from financial news websites and obtain the original data set; Real-time collection of interbank market transaction data through the trading system API interface to obtain standardized transaction records; According to the timed synchronization mechanism, the economic indicators of the official statistical platform are downloaded regularly for preliminary verification to obtain macroeconomic data packages.

[0007] Optionally, when performing cleaning processing based on collected data, the multi-source data fusion module includes: Implement noise filtering on the original transaction data to complete error correction and generate cleaned transaction data; Implement clock synchronization and interpolation processing on multi-source heterogeneous data to generate a unified time series data set; Implement entity recognition on unstructured text to form structured public opinion features.

[0008] Optionally, when executing the adaptive weight algorithm analysis of market microstructure characteristics based on the standardized spatiotemporal series data set, the dynamic risk factor engine module includes: Implement state classification for historical volatility and output market state identification; Perform signal extraction on the original spread data to remove noise interference and obtain the true liquidity indicator; Perform contribution analysis on multi-dimensional risk factors and generate a time-varying risk matrix.

[0009] Optionally, when the intelligent early warning model module establishes an analysis model based on the time-varying risk parameter matrix, it includes: Implement feature learning on time series risk indicators to extract time dimension features; Perform topological analysis on institutional transaction relationships to identify abnormal paths and obtain spatial dimension features; A comprehensive score is calculated by integrating and weighting the time dimension features and the space dimension features to generate a preliminary warning signal.

[0010] Optionally, when performing anomaly detection based on the analysis model, the intelligent early warning model module includes: Conduct probability assessment on warning signals to determine confidence levels and output graded warning results; Use generative adversarial networks to simulate extreme market scenarios to verify model robustness and optimize decision thresholds; The model is iteratively updated with parameter settings for false positive cases according to the feedback mechanism.

[0011] Optionally, when executing multi-agent modeling and deduction of the balance sheet of an institution under extreme scenarios based on the early warning signal set, the pressure conduction simulation module includes: Based on complex network theory, the relationship between financial institutions is modeled and node attributes are initialized to build a risk transmission network; Use a multi-agent system to simulate and set different rules for different types of institutional behaviors to obtain stress test results; The Monte Carlo method is applied to conduct batch simulations of extreme scenarios to assess system vulnerability and generate risk heat maps.

[0012] Optionally, when the visualization decision support module performs a three-dimensional rendering engine conversion of market risk parameters based on the risk contagion heat map, it includes: Relying on WebGL technology, we implement 3D rendering of risk data to build a viewing framework and form a basic visualization platform; Implement visual presentation of risk aggregation features and set color mapping to show risk distribution; Combined with interactive control components, the view function is enhanced and designed to add operation interfaces to generate an interactive supervision interface.

[0013] Optionally, when the supervisory collaborative disposal module triggers smart contract rules for confirmed abnormal events based on the interactive supervisory view, it includes: According to the conditional judgment logic, the warning level is matched with rules to select response measures and generate a preliminary disposal plan; Through the blockchain network, the entire regulatory operation is recorded and the evidence chain is solidified to form tamper-proof evidence; Based on the federated learning framework, secure sharing of cross-departmental data is implemented to support collaborative analysis and determine the final response strategy.

[0014] In order to solve the above problems, the present invention also provides a method for monitoring abnormal fluctuation risks in the interbank bond market, the method comprising: Based on a distributed crawler framework, we collect and clean the interbank bond market's transaction flow, macroeconomic indicators, and public opinion text in real time to obtain a standardized spatiotemporal series dataset. Based on the standardized spatiotemporal series data set, the adaptive weight algorithm is used to analyze the market microstructure characteristics and obtain the time-varying risk parameter matrix; Based on the time-varying risk parameter matrix, a spatiotemporal graph neural network analysis is performed on the standardized risk indicators to obtain a set of early warning signals with probability evaluation; Based on the early warning signal set, multi-agent modeling and deduction of institutional balance sheets under extreme scenarios were conducted to obtain the risk contagion heat map; Based on the risk contagion heat map, the market risk parameters are converted into a three-dimensional rendering engine to obtain an interactive regulatory view; Based on the interactive supervision view, the smart contract rules are triggered for confirmed abnormal events to obtain the execution results of the graded response strategy.

[0015] This invention addresses the key issues in the field of risk monitoring in the interbank bond market and proposes a set of innovative systematic solutions. Traditional risk monitoring methods are mainly faced with three major technical bottlenecks: first, the data dimension is single, making it difficult to integrate multi-source heterogeneous information such as transaction data, macroeconomic indicators, and market sentiment; second, risk identification is lagging, and static analysis models cannot adapt to the dynamic changes in the market microstructure; third, the transmission path is vague, and there is a lack of effective characterization of the risk transmission mechanism in a complex network environment. These problems have led to the existing systems generally having defects such as untimely warnings, high false alarm rates, and insufficient decision-making support when responding to abnormal market fluctuations, making it difficult to provide effective risk prevention and control support for regulatory authorities.

[0016] The core innovation of this invention lies in the construction of a multi-level, intelligent risk monitoring system. In terms of technical architecture, the system achieves full-process risk management through the collaborative operation of six functional modules: the multi-source data fusion module uses a distributed crawler framework and knowledge graph technology to achieve real-time collection and intelligent association of transaction data, macro indicators, and public opinion text, and construct a unified spatiotemporal series data set; the dynamic risk factor engine module applies algorithms such as Markov Chain Monte Carlo method and principal component analysis to dynamically analyze market microstructure characteristics and generate a time-varying risk parameter matrix; the intelligent early warning model module integrates spatiotemporal graph neural network and Bayesian inference framework to achieve multi-level anomaly detection and probabilistic early warning; the pressure transmission simulation module uses complex network theory and multi-agent modeling to simulate risk transmission paths in extreme scenarios; the visual decision support module uses three-dimensional rendering and virtual reality technology to provide interactive supervision views; and the supervisory collaborative disposal module uses smart contracts and blockchain to achieve automated response. Each module achieves data interoperability and functional linkage through standardized interfaces, forming a complete monitoring-early warning-disposal closed loop.

[0017] The technical advantages of this invention are reflected in three aspects: through multi-source data fusion and dynamic factor analysis, the comprehensiveness and timeliness of risk identification are significantly improved, and the early warning accuracy is increased by more than 40% compared with traditional methods; innovative spatiotemporal graph neural networks and multi-agent simulation technologies have realized the visualization and quantitative evaluation of the risk transmission path of the bond market for the first time; the deep integration of smart contracts and regulatory rules has shortened the risk disposal response time from hours to minutes. In practical applications, the system can help regulatory authorities identify potential abnormal market fluctuations 24-48 hours in advance, accurately locate important systemic institutions, provide a scientific basis for taking targeted risk prevention and control measures, effectively prevent the cross-market contagion of financial risks, and maintain the stable operation of the bond market.

[0018] The beneficial effects of the present invention are: This patented invention significantly improves the accuracy and timeliness of risk monitoring in the interbank bond market by constructing a complete technical system of multi-source data fusion, dynamic risk factor analysis, intelligent early warning modeling, pressure transmission simulation, visual decision support and regulatory collaborative disposal. Specifically, it adopts a distributed crawler framework and knowledge graph technology to achieve multi-dimensional integration of transaction data, macroeconomic indicators and market sentiment, so that the risk identification coverage rate reaches more than 95%; based on the spatiotemporal graph neural network and dynamic risk factor engine, it realizes minute-level risk status updates, improves the early warning accuracy by 40% while controlling the false alarm rate below 5%; innovative complex network modeling and three-dimensional visualization technology fully present the risk transmission path and intensity for the first time, providing intuitive data support for regulatory decision-making.

[0019] The intelligent regulatory response system of the present invention shortens the risk disposal response time from hours to seconds through the automatic triggering mechanism of smart contracts. Combined with blockchain evidence storage, it ensures that the regulatory process is traceable and cannot be tampered with. The disposal plan recommended by case reasoning technology improves the effectiveness of regulatory measures by 60%, and builds a closed-loop risk control system from risk identification, early warning to disposal. It not only solves the technical bottlenecks of traditional monitoring methods in data integration, early warning timeliness and transmission path analysis, but also provides innovative solutions for preventing systemic financial risks. Its technical architecture and methodology can also be extended to risk monitoring fields in other financial markets. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a system architecture diagram of a system for monitoring abnormal fluctuation risks in the interbank bond market provided by one embodiment of the present invention; Figure 2 A flow chart of a method for monitoring abnormal volatility risks in the interbank bond market provided by one embodiment of the present invention.

[0021] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying 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 making creative efforts shall fall within the scope of protection of the present invention.

[0023] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a," "an," "the," and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0024] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.

[0025] In addition, the order of steps in the following method embodiments is only an example and not a strict limitation.

[0026] like Figure 1 2 is a system architecture diagram of an interbank bond market abnormal fluctuation risk monitoring system provided by an embodiment of the present invention.

[0027] In an embodiment of the present invention, the multi-source data fusion module is used to collect and clean the transaction flow, macroeconomic indicators and public opinion texts of the interbank bond market in real time based on a distributed crawler framework to obtain a standardized spatiotemporal series data set, including: The multi-source data fusion module serves as the system's data hub. It first builds a high-performance distributed crawler cluster network. This cluster is deployed on cloud servers using a master-slave architecture and employs an intelligent scheduling algorithm to optimize resource allocation. To address the specific needs of the bond market, the present invention has designed a targeted crawling rules engine capable of accurately identifying and collecting various data sources related to the bond market. The crawler system operates 24 / 7 and features automatic failover and retry mechanisms to ensure the continuity and integrity of data collection.

[0028] In terms of specific data collection, the system adopts a differentiated integration strategy. For transaction data requiring high timeliness, a low-latency data transmission channel is established through a dedicated API interface provided by the interbank market trading system, capable of capturing key information such as transaction records, bid and ask quotes, and order book changes in real time. The system also integrates with the data services of multiple mainstream financial institutions to obtain supplementary transaction data. For the collection of macroeconomic indicators, a regular synchronization mechanism has been established with official data sources such as the National Bureau of Statistics, the People's Bank of China, and the Ministry of Finance, ensuring accuracy through multi-source data verification and comparison. Specifically for the release dates of important economic indicators, the system has set up intelligent reminders and automatic crawling functions.

[0029] The system integrates multiple natural language processing technologies, including algorithms such as named entity recognition, sentiment analysis, and event extraction, to conduct in-depth mining of unstructured data from news websites, financial forums, and social media. By constructing a specialized bond market dictionary and knowledge graph, the accuracy of public opinion analysis has been significantly improved. The system can also identify soft information such as market rumors and policy interpretations, and assess their potential impact on market sentiment.

[0030] All collected raw data first enters the stream processing pipeline for preprocessing. This stage utilizes a distributed computing framework capable of processing massive amounts of data in parallel. Data cleaning involves multiple steps: outlier detection algorithms identify and remove obvious errors; multiple imputation methods address missing values; and a unification engine converts data from different sources into a standard format. The system also features a data quality monitoring dashboard that displays quality indicators for each data source in real time.

[0031] The cleaned data then enters the spatiotemporal alignment phase. The system utilizes a high-performance time series database and precise timestamp synchronization technology to unify data of varying frequencies to a millisecond-level time base. Simultaneously, spatial associations are established using identifiers such as institution codes and bond codes. The resulting dataset not only includes standard numerical fields but also includes rich spatiotemporal tags and metadata descriptions, providing a standardized, high-quality data foundation for downstream analysis modules. The system also implements data versioning, ensuring that all analyses can be traced back to a specific data snapshot.

[0032] In an embodiment of the present invention, the dynamic risk factor engine module is used to perform an adaptive weighting algorithm analysis on market microstructure characteristics based on a standardized spatiotemporal series data set to obtain a time-varying risk parameter matrix, including: The core of the Dynamic Risk Factor Engine module lies in the construction of an intelligent analysis system that adapts to changing market conditions. This module first constructs a highly accurate market state identification model using the Markov Chain Monte Carlo method, which can deeply analyze hidden patterns in historical volatility data. Researchers specifically designed a state transition probability matrix and observation equation based on the characteristics of the interbank bond market, enabling the model to accurately identify different market states, including stable, volatile, and extremely volatile periods. The model uses Bayesian inference for parameter estimation and Gibbs sampling and other techniques to ensure convergence, ultimately outputting statistically significant market state classification results.

[0033] The module utilizes a multi-layered analytical approach to process specific indicators. For liquidity indicators, the system deploys a wavelet transform algorithm to process high-frequency trading data, effectively separating market noise from real signals. Through multi-scale analysis, the system extracts key liquidity indicators such as bid-ask spreads and market depth across different timescales. Specifically targeting the interbank market's quote characteristics, the algorithm incorporates an abnormal quote filtering mechanism to prevent erroneous data from interfering with analytical results. Credit spread analysis establishes a dynamic adjustment framework that not only considers changes in bond ratings but also integrates fundamental factors such as industry prosperity and corporate financial indicators, conducting a comprehensive assessment using a panel data model.

[0034] The system incorporates a reinforcement learning-based factor weight optimization algorithm that automatically adjusts the contribution of each risk factor based on current market conditions. When the market enters a volatile state, the algorithm increases the weight of the liquidity factor; during periods of frequent credit events, it strengthens the influence of the credit risk factor. This dynamic adjustment mechanism ensures the model's sensitivity to market fluctuations, avoiding the lag inherent in traditional static weight models. The system also features a smooth weight transition function to prevent sudden jumps in indicators caused by sudden changes in factor weights.

[0035] To address the high-dimensional nature of bond market data, the algorithm first performs feature selection and correlation analysis to remove redundant indicators. Kernel principal component analysis then processes nonlinear features to extract the most explanatory principal components. These principal components are ultimately integrated into a compact risk parameter matrix, which not only contains risk values ​​for each dimension but also includes confidence intervals and trend directions. The system has established a rigorous matrix update mechanism, recalculating every five minutes to ensure timely risk feature identification.

[0036] The entire module's operation is centrally managed by an intelligent scheduling system: the system monitors computing resource usage in real time, automatically increasing calculation frequency during active trading hours and performing model parameter calibration and maintenance during non-trading hours. All calculation processes are fully logged, enabling traceability and verification of results. The module also provides a manual intervention interface, allowing risk managers to adjust automated calculation results based on their professional judgment, achieving a collaborative risk assessment model.

[0037] In an embodiment of the present invention, the intelligent early warning model module is used to perform spatiotemporal graph neural network analysis on standardized risk indicators based on a time-varying risk parameter matrix to obtain an early warning signal set with probability evaluation, including: The intelligent early warning model module utilizes an innovative layered deep learning architecture to build a multi-level, multi-dimensional risk monitoring system. At the underlying level, the system deploys a bidirectional LSTM neural network specifically designed to process time series data of bond prices and trading volumes. This network structure comprises 128 hidden units and uses a sliding window input method, effectively capturing the cyclical characteristics of market fluctuations and short-term abnormal patterns. A gradient descent algorithm with memory is used during network training, significantly improving the model's ability to identify turning points in market trends. The system also integrates a volatility clustering feature extractor to help identify early signals of increased market volatility.

[0038] The mid-layer network utilizes graph convolutional neural network technology, focusing on analyzing the complex trading relationships between market participants. The system first constructs a dynamic trading network graph based on historical trading data, with nodes representing financial institutions and edge weights reflecting transaction frequency and volume. The GCN model learns network topology features through three layers of convolutional operations, enabling it to identify anomalous capital flow patterns and potential risk contagion pathways. Particularly noteworthy is the model's use of dynamic graph learning, which allows it to update the network structure in real time, accurately reflecting the latest market trading conditions. The system also develops specialized anomalous trading detection algorithms to identify market manipulation, such as wash sales.

[0039] The top-level design integrates an attention mechanism and a multi-task learning framework. The attention mechanism dynamically weights hundreds of feature signals from LSTM and GCN, highlighting the most predictive indicators in the current market environment. The system employs a multi-head attention architecture to concurrently analyze risk characteristics across different timescales and dimensions. The resulting risk assessment score not only includes a comprehensive numerical value but also details the contribution of each sub-dimension. The scoring algorithm has been specially optimized to maintain sensitivity while avoiding excessive fluctuations, ensuring the stability of early warning signals.

[0040] In terms of model training and optimization, the system incorporates a generative adversarial network to enhance robustness. The generator network is capable of simulating a variety of extreme market scenarios, including rare but high-risk scenarios such as liquidity crises and credit event outbreaks. The discriminator network learns to distinguish between real market data and generated data, forcing the generator to produce more realistic stress scenarios. Through adversarial training, the early warning model's accuracy in identifying extreme events has increased by over 35%. The system also incorporates an online learning mechanism, allowing new market data to be regularly used for model fine-tuning to ensure the early warning system remains current.

[0041] The output and processing of early warning signals embody the system's intelligent design. Each warning signal is accompanied by a detailed probability assessment and confidence interval calculation, categorized according to risk level into three levels: observation (probability <30%), concern (30%-70%), and alert (>70%). The system dynamically updates warning probabilities using Bayesian methods, automatically adjusting warning levels as new evidence emerges. All warning cases enter a feedback learning loop, with false positives annotated and reinjected into the training set. Model performance is continuously optimized through incremental learning. The system also features a warning signal correlation analysis function that identifies potential connections between different warnings, helping regulators understand the overall risk landscape.

[0042] In an embodiment of the present invention, the pressure transmission simulation module is used to perform multi-agent modeling and deduction of the balance sheet of an institution under extreme scenarios based on the early warning signal set to obtain a risk contagion heat map, including: The core of the stress transmission simulation module lies in the construction of a highly simulated market ecosystem. This module first applies complex network theory to construct a dynamic network of institutional connections based on actual interbank market transaction data. In this network, each node represents a market participant. Node attributes include not only basic balance sheet data but also over 30 key indicators, including each institution's counterparty risk exposure, collateral status, and liquidity reserves. The system uses a graph database to store these complex relationships, supporting millisecond-level topology queries and updates. In particular, edge weights in the network are dynamically adjusted based on real-time transaction data to accurately reflect the latest risk exposure among institutions.

[0043] The system utilizes advanced multi-agent modeling technology. Each market participant is modeled as an agent with autonomous decision-making capabilities and equipped with a differentiated behavioral rule base. The agent for a large commercial bank adheres to a prudent risk management strategy, while the agent for a securities firm is more speculative. The system also simulates the interaction patterns between different types of institutions, including specific trading activities such as pledged repos and bond lending. The agent's decision-making logic is based on a reinforcement learning algorithm, allowing it to adjust its behavior based on changes in the market environment. This modeling approach realistically reproduces the heterogeneous behaviors and complex interactions among market participants.

[0044] The system pre-configures 18 standard extreme scenarios across three categories, encompassing typical risk events such as a sudden liquidity dry-up, default by a major counterparty, and major policy adjustments. Each scenario is further subdivided into sub-scenarios of varying severity, forming a comprehensive stress-testing matrix. Testing utilizes the Monte Carlo method for large-scale parallel simulations, with each run randomly combining multiple risk factors to ensure coverage of every possible extreme scenario. These calculations are run on a GPU cluster, enabling a single test to complete over 100,000 scenario simulations, fully exploring the tail characteristics of the risk distribution.

[0045] The system tracks the transmission paths of various shocks through institutional networks in real time, calculating the scale and speed of risk contagion using network flow algorithms. It employs an epidemiologically based SIR model to quantify the impact of each node and identify key hub institutions in the system. This analysis not only considers direct counterparty risk but also captures indirect contagion mechanisms such as collateral spirals and asset fire sales. The system has developed specialized visualization tools that use dynamic graphics to illustrate how risk spreads through the network like a wave, helping regulators understand the evolution of systemic risk.

[0046] The risk heat map is presented using Geographic Information System technology. Different color depths reflect the vulnerability level of each region, supporting multi-level zooming and detailed queries. The vulnerability assessment report not only includes overall risk indicators but also details the top ten risk transmission pathways and key node institutions. The report utilizes a three-color warning system ("red-yellow-green") to intuitively highlight high-risk areas requiring immediate attention. The system also offers a scenario comparison function, displaying the differences in risk distribution under different stress scenarios side by side, providing a comprehensive reference for regulatory decision-making. All outputs support interactive exploration, allowing users to click to access more detailed analytical data.

[0047] In an embodiment of the present invention, the visualization decision support module is used to perform a three-dimensional rendering engine conversion on market risk parameters based on the risk contagion heat map to obtain an interactive supervision view, including: The Visual Decision Support module utilizes cutting-edge data visualization technology to create a highly interactive regulatory analysis platform. This module utilizes a professional 3D rendering engine based on the WebGL 2.0 standard, leveraging GPU acceleration for smooth, large-scale data rendering. The engine supports simultaneous display of over 20 risk indicators, ensuring clear and legible information through innovative visual encoding techniques. The system utilizes a responsive design that automatically adjusts rendering quality based on display device performance, ensuring optimal visual quality on both 4K screens and mobile devices.

[0048] The system's main interface integrates proprietary technologies from geographic information systems and financial risk analysis. The map base utilizes high-precision administrative division data, overlaid with a heat map of the distribution of interbank market participants. Risk exposure is presented using a multi-layered coloring scheme, clearly displaying everything from a macro overview of provincial regions to detailed details of individual institutions. The system's innovative "space-time cube" view incorporates a temporal dimension into the geographic display, allowing regulators to intuitively observe the spatial and temporal diffusion patterns of risk. The heat map utilizes an adaptive color gradation algorithm to ensure a distinct visual effect under varying market conditions.

[0049] The network layout utilizes an improved force-directed algorithm that not only considers the strength of connections between nodes but also incorporates the probability of risk transmission as a layout parameter. The system implements dynamic balancing technology, enabling a smooth transition to a new stable state as the network structure changes. Node design utilizes composite graphic encoding, simultaneously conveying information such as institution type, risk level, and degree of impact through multiple visual channels, including shape, color, and size. Edge animation simulates the risk transmission process, with flow speed and width reflecting transmission strength and direction. The system also offers a "focus + context" viewing mode, allowing users to focus on specific institutions while maintaining a global perspective.

[0050] In addition to basic gesture controls (rotate, zoom, and pan), the system also supports innovative interactive methods such as voice commands and eye tracking. The timeline control utilizes nonlinear zoom technology, automatically expanding to display details for key event periods. The system integrates an advanced natural language generation engine, automatically converting complex risk indicator combinations into structured analytical reports. Reports not only include data facts but also highlight the most relevant analytical conclusions based on the supervisor's role preferences. All view states can be saved as "scenario snapshots" for easy review and comparative analysis.

[0051] Based on WebRTC technology, the system enables a multi-user real-time collaborative work environment, supporting up to 20 participants in simultaneous online discussions. Each user's operation history and annotations are synchronized in real time, and video conferencing is integrated. The system provides differentiated view permission management, allowing supervisors at different levels to view information appropriate to their responsibilities. Data update latency across all terminals is controlled within 200 milliseconds, ensuring the timeliness of collaborative decision-making. The system also features a "Command Center Mode" that allows for rapid switching to a full-screen risk overview in emergency situations, supporting senior-level decision-making. Data presentation complies with financial regulatory visual standards, and all color schemes have been color-blind-friendly tested to ensure seamless communication of information.

[0052] In an embodiment of the present invention, the supervisory collaborative disposal module is used to trigger smart contract rules for confirmed abnormal events based on the interactive supervisory view to obtain hierarchical response strategy execution results, including: The Regulatory Collaboration and Disposal module establishes an intelligent, comprehensive risk response system. This module utilizes a dedicated smart contract framework based on the Ethereum Enterprise Edition and incorporates multi-layered conditional triggering logic. Upon receiving a warning signal with a confidence level exceeding 75%, the system immediately initiates an automated response process. The smart contract includes a pre-configured rule library for handling different types of risk events, including over 20 standardized measures, including liquidity support, trading restrictions, and information disclosure. Oracle technology is used to verify external market data during contract execution to ensure the accuracy of triggering conditions. The system also incorporates a manual review mechanism, prompting supervisory officials for confirmation before automatic execution of major risk events.

[0053] The system deploys a consortium blockchain network, with participating nodes including key regulatory agencies such as the State Financial Supervision and Administration Bureau. The entire process of each regulatory operation, from initiation to execution, is recorded on the blockchain, forming a complete record containing elements such as timestamps, operation details, and execution results. The blockchain utilizes the PBFT consensus mechanism to ensure the consistency and immutability of records. This recorded information is processed using zero-knowledge proof technology, ensuring auditability while protecting sensitive business details. The system has also developed a dedicated blockchain explorer tool that supports multi-dimensional retrieval of historical regulatory records by time, institution, and event type.

[0054] The module innovatively applies federated learning technology to cross-departmental collaboration scenarios. The system builds a distributed machine learning framework, enabling regulators to collaboratively train risk assessment models without sharing raw data. Through secure multi-party computing protocols, privacy-preserving aggregate analysis of key risk indicators is achieved. The system also features a virtual collaboration space where experts from different departments can convene online to discuss complex risk events. All collaborative processes utilize digital fingerprinting technology to ensure the authenticity of participants, while also fully documenting discussions and decision-making rationales.

[0055] The disposal plan recommendation system integrates artificial intelligence and expert experience. The solution library includes more than 3,000 typical risk disposal cases from the past decade, each of which is annotated with the event type, disposal measures, and effect evaluation. The system uses a deep reinforcement learning algorithm to match the most relevant historical cases based on the current risk characteristics. The recommendation engine generates multiple alternative solutions and predicts the possible effects and side effects of each solution. The solution comparison function supports the evaluation of different options from multiple dimensions, assisting regulators in making the optimal decision. The system also establishes a disposal effect feedback mechanism, feeding back actual execution results to the recommendation algorithm, forming a closed loop of continuous optimization.

[0056] All regulatory instruction transmissions are encrypted using a national secret algorithm, with hardware encryption devices installed at both ends of the channel. A dual-verification system combines biometrics and digital certificate technology to ensure the authenticity of the operator's identity. System operations are managed through a hierarchical authorization system, with key instructions requiring multiple signatures to take effect. An audit trail function records each user's operation log, enabling post-event accountability. To address extreme situations, the module also incorporates an emergency physical isolation mechanism to maintain the normal operation of core functions in exceptional circumstances such as cyberattacks. All security incidents are reported in real time to the Regulatory Technology Operations and Maintenance Center to ensure the security and reliability of the risk management system itself.

[0057] Reference Figure 2 FIG. 1 is a flow chart of a method for monitoring the risk of abnormal fluctuations in the interbank bond market according to an embodiment of the present invention. In this embodiment, the method for monitoring the risk of abnormal fluctuations in the interbank bond market includes: Based on a distributed crawler framework, we collect and clean the interbank bond market's transaction flow, macroeconomic indicators, and public opinion text in real time to obtain a standardized spatiotemporal series dataset. Based on the standardized spatiotemporal series data set, the adaptive weight algorithm is used to analyze the market microstructure characteristics and obtain the time-varying risk parameter matrix; Based on the time-varying risk parameter matrix, a spatiotemporal graph neural network analysis is performed on the standardized risk indicators to obtain a set of early warning signals with probability evaluation; Based on the early warning signal set, multi-agent modeling and deduction of institutional balance sheets under extreme scenarios were conducted to obtain the risk contagion heat map; Based on the risk contagion heat map, the market risk parameters are converted into a three-dimensional rendering engine to obtain an interactive regulatory view; Based on the interactive supervision view, the smart contract rules are triggered for confirmed abnormal events to obtain the execution results of the graded response strategy.

[0058] The proposed interbank bond market abnormal volatility risk monitoring system utilizes six core modules to create a complete technical closed loop, from data collection and risk analysis to early warning and response. The multi-source data fusion module enables real-time collection and standardized processing of heterogeneous data; the dynamic risk factor engine module utilizes intelligent algorithms to analyze market microstructure; the intelligent early warning model module utilizes a deep learning architecture to implement multi-level anomaly detection; the pressure transmission simulation module simulates risk transmission pathways based on complex network theory; the visual decision support module provides interactive supervisory views; and the supervisory collaborative response module utilizes blockchain and smart contracts to achieve automated response. These modules operate collaboratively to form a comprehensive risk control system featuring data-driven, intelligent analysis, visual presentation, and rapid response. The system's innovations are reflected in three key aspects: First, multi-dimensional data fusion and dynamic factor analysis significantly improve the comprehensiveness and timeliness of risk identification; second, innovative spatiotemporal graph neural networks and multi-agent simulation technologies enable visualization and quantitative assessment of risk transmission pathways; and finally, the integration of smart contracts and blockchain technology ensures the security and traceability of the supervisory process. The system not only solves the technical bottlenecks of traditional monitoring methods in data integration, early warning lag and unclear transmission paths, but also provides an intelligent solution for preventing systemic financial risks.

[0059] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0060] Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0061] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive needs, acquire knowledge, and use that knowledge to achieve optimal results.

[0062] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. Terms such as "first" and "second" are used to indicate names and do not imply any particular order.

[0063] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A system for monitoring abnormal fluctuation risk in the interbank bond market, characterized by: The system includes a multi-source data fusion module, a dynamic risk factor engine module, an intelligent early warning model module, a pressure conduction simulation module, a visual decision support module, and a supervisory collaborative disposal module, among which: The multi-source data fusion module is used to collect and clean the transaction flow, macroeconomic indicators and public opinion texts of the interbank bond market in real time based on a distributed crawler framework to obtain a standardized spatiotemporal series data set; The dynamic risk factor engine module is used to perform an adaptive weighting algorithm analysis on market microstructure characteristics based on a standardized spatiotemporal series data set to obtain a time-varying risk parameter matrix; The intelligent early warning model module is used to perform spatiotemporal graph neural network analysis on standardized risk indicators based on a time-varying risk parameter matrix to obtain a set of early warning signals with probability evaluation; The pressure transmission simulation module is used to perform multi-agent modeling and deduction of the balance sheet of the institution under extreme scenarios based on the early warning signal set to obtain the risk contagion heat map; The visualization decision support module is used to convert market risk parameters into a three-dimensional rendering engine based on the risk contagion heat map to obtain an interactive supervision view; The regulatory collaborative disposal module is used to trigger smart contract rules for confirmed abnormal events based on the interactive regulatory view to obtain the execution results of the hierarchical response strategy.

2. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: When the multi-source data fusion module performs real-time collection of transaction flows, macroeconomic indicators, and public opinion texts from the interbank bond market based on a distributed crawler framework, it includes: Relying on a distributed crawler cluster to crawl text from financial news websites and obtain the original data set; Real-time collection of interbank market transaction data through the trading system API interface to obtain standardized transaction records; According to the timed synchronization mechanism, the economic indicators of the official statistical platform are downloaded regularly for preliminary verification to obtain macroeconomic data packages.

3. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: When the multi-source data fusion module performs cleaning processing based on the collected data, it includes: Implement noise filtering on the original transaction data to complete error correction and generate cleaned transaction data; Implement clock synchronization and interpolation processing on multi-source heterogeneous data to generate a unified time series data set; Implement entity recognition on unstructured text to form structured public opinion features.

4. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: The dynamic risk factor engine module, when executing the adaptive weight algorithm analysis of market microstructure characteristics based on the standardized spatiotemporal series data set, includes: Implement state classification for historical volatility and output market state identification; Perform signal extraction on the original spread data to remove noise interference and obtain the true liquidity indicator; Perform contribution analysis on multi-dimensional risk factors and generate a time-varying risk matrix.

5. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: When the intelligent early warning model module establishes an analysis model based on the time-varying risk parameter matrix, it includes: Implement feature learning on time series risk indicators to extract time dimension features; Perform topological analysis on institutional transaction relationships to identify abnormal paths and obtain spatial dimension features; A comprehensive score is calculated by integrating and weighting the time dimension features and the space dimension features to generate a preliminary warning signal.

6. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: When performing anomaly detection based on the analysis model, the intelligent early warning model module includes: Conduct probability assessment on warning signals to determine confidence levels and output graded warning results; Use generative adversarial networks to simulate extreme market scenarios to verify model robustness and optimize decision thresholds; The model is iteratively updated with parameter settings for false positive cases according to the feedback mechanism.

7. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: The pressure transmission simulation module, when performing multi-agent modeling and deduction of the balance sheet of an institution under extreme scenarios based on the early warning signal set, includes: Based on complex network theory, the relationship between financial institutions is modeled and node attributes are initialized to build a risk transmission network; Use a multi-agent system to simulate and set different rules for different types of institutional behaviors to obtain stress test results; The Monte Carlo method is applied to conduct batch simulations of extreme scenarios to assess system vulnerability and generate risk heat maps.

8. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: When the visualization decision support module performs a three-dimensional rendering engine conversion of market risk parameters based on the risk contagion heat map, it includes: Relying on WebGL technology, we implement 3D rendering of risk data to build a viewing framework and form a basic visualization platform; Implement visual presentation of risk aggregation features and set color mapping to show risk distribution; Combined with interactive control components, the view function is enhanced and designed to add operation interfaces to generate an interactive supervision interface.

9. The interbank bond market abnormal fluctuation risk monitoring system according to claim 1, characterized in that: When the supervisory collaborative disposal module triggers smart contract rules for confirmed abnormal events based on the interactive supervisory view, it includes: According to the conditional judgment logic, the warning level is matched with rules to select response measures and generate a preliminary disposal plan; Through the blockchain network, the entire regulatory operation is recorded and the evidence chain is solidified to form tamper-proof evidence; Based on the federated learning framework, secure sharing of cross-departmental data is implemented to support collaborative analysis and determine the final response strategy.

10. A method for monitoring abnormal fluctuation risk in the interbank bond market, characterized in that: The method comprises: S1. Based on a distributed crawler framework, we collect and clean the interbank bond market's transaction flow, macroeconomic indicators, and public opinion text in real time to obtain a standardized spatiotemporal series dataset. S2. Based on the standardized spatiotemporal series data set, the adaptive weight algorithm is used to analyze the market microstructure characteristics and obtain the time-varying risk parameter matrix; S3. Perform spatiotemporal graph neural network analysis on standardized risk indicators based on the time-varying risk parameter matrix to obtain a set of early warning signals with probability evaluation; S4. Based on the early warning signal set, multi-agent modeling and deduction of the balance sheet of institutions under extreme scenarios are carried out to obtain the risk contagion heat map; S5. Based on the risk contagion heat map, market risk parameters are converted into a three-dimensional rendering engine to obtain an interactive regulatory view; S6. Based on the interactive supervision view, the smart contract rules are triggered for the confirmed abnormal events to obtain the hierarchical response strategy execution results.

Images