Intelligent risk analysis and research system and method based on browser plug-in

Abstract

The invention discloses an intelligent risk analysis and research system and method based on a browser plug-in, and relates to the technical field of financial science and technology. The system comprises a visual process arrangement interface, a process engine and an AI model. The system receives a modular risk control node process arranged by a user through a visual interface, and is driven and executed by a process engine. The core of the method is that in the process of process execution, a system can respond to an interrupt waiting node or a user natural language instruction to suspend the process and wait for interaction; the instruction intention can be analyzed to realize flow skipping or backtracking; after the nodes are executed again, the AI model carries out comparative analysis on execution results and generates a report; and finally integrating node outputs to generate a structured risk control conclusion. According to the method, the problems of low efficiency, lack of deep insight and flexibility of a traditional risk control process are solved, intelligent risk analysis of man-machine cooperation is realized, and the method is particularly suitable for financial risk management scenes such as commercial factoring and the like.

Description

Technical Field

[0001] This invention relates to the intersection of financial technology and information technology, and in particular to an intelligent risk analysis research system and method based on a browser plugin. Background Technology

[0002] With the rapid development of commercial factoring, efficient and accurate risk assessment of financing companies has become a core requirement of the industry. Currently, the risk control process of commercial factoring companies generally relies heavily on manual operation. Typically, risk control personnel need to manually switch frequently between different platforms (such as enterprise information query platforms, legal litigation websites, tax systems, internal databases, etc.) within a single workday to perform information queries, data entry, cross-validation, and final report writing. This model is essentially an assembly line with risk control personnel as the "central processing unit (CPU)."

[0003] Existing technical solutions suffer from the following inherent and insurmountable drawbacks: Manual processing is slow, unable to handle massive amounts of multi-source, unstructured enterprise information, and prone to oversights and errors in repetitive operations, becoming a bottleneck for business expansion; traditional methods often remain at the level of listing information and simple indicator calculations, lacking the ability for in-depth data mining and correlation. Risk control personnel struggle to conduct multi-dimensional, non-linear correlation analysis of information such as the company's equity structure, supply chain relationships, financial data, and legal proceedings, resulting in superficial risk assessment conclusions that fail to identify complex, potential associated risks. Summary of the Invention

[0004] In view of the above-mentioned shortcomings in the current field of financial technology and information technology intersection, the present invention provides an intelligent risk analysis research system and method based on browser plugins. This technical solution can fundamentally change the status quo, freeing risk control personnel from tedious repetitive work and combining their professional experience with the powerful data processing and intelligent analysis capabilities of machines, thereby achieving the effect of transforming risk assessment from a "labor-intensive" to an "intelligent-driven" paradigm.

[0005] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions:

[0006] A browser plugin-based intelligent risk analysis and research system, the system comprising:

[0007] A visual workflow orchestration interface, loaded in a browser, is used to receive workflow orchestrations of modular risk control nodes by users through drag-and-drop and connection operations;

[0008] A process engine is used to execute the analysis process consisting of the risk control nodes;

[0009] The AI ​​model is coupled with the process engine and risk control nodes.

[0010] The system is configured as follows:

[0011] During process execution, in response to reaching a preset interruption waiting node or receiving a natural language command from the user, the current process is paused and the user is waiting for interaction, which may include uploading files or providing supplementary information.

[0012] The intent of the natural language instruction is parsed, and the process jumps to the target node in the flow to continue execution or backtracks to the previous node to re-execute, based on the intent.

[0013] After the node is re-executed, the AI ​​model compares and analyzes the execution results of the node in different contexts and generates a comparison report.

[0014] After the process is completed, the outputs of each node are integrated to generate a structured risk analysis conclusion.

[0015] According to one aspect of the present invention, the visual process orchestration interface is further configured to, in response to user operation, present a visual process orchestration and execution interface on the right side of the browser, which can be manually enlarged to full screen.

[0016] According to one aspect of the invention, the process engine supports at least two process construction modes:

[0017] AI-assisted orchestration mode: In response to the analysis goals set by the user, the AI ​​model recommends and automatically adds relevant risk control nodes to the orchestration interface;

[0018] AI Fully Autonomous Mode: In response to the company name entered by the user, the AI ​​model autonomously constructs and executes a complete analysis process.

[0019] According to one aspect of the present invention, the modular risk control node includes at least one of the following:

[0020] The enterprise panoramic insight node is configured to acquire enterprise information and build a relational graph through one or more of the following methods: API interface, web crawler, RPA robotic process automation, or multimodal AI analysis.

[0021] The multi-dimensional financial statement deep diagnostic node is configured to perform cross-period and cross-account linkage analysis on the company's balance sheet, income statement, and cash flow statement, and to perform counterintuitive reasoning and stress testing.

[0022] The dynamic file knowledge base node is configured to allow users to upload and analyze related files, and the AI ​​model performs multimodal understanding, weight learning, and evidence conflict resolution on the file content.

[0023] The intelligent analysis node for judicial litigation is configured to use natural language processing technology to analyze judgment documents and conduct in-depth analysis of the cause of action and assessment of the litigation impact.

[0024] The supply chain resilience analysis node is configured to construct a supply chain relationship graph of the target enterprise by the AI ​​model, inject multi-dimensional risk data, and run stress test scenarios simulating external shocks.

[0025] According to one aspect of the present invention, the enterprise panoramic insight node is further configured to: acquire data using an intelligent hierarchical strategy, prioritize calling API interfaces, enable web crawlers when API interface calls fail, enable RPA simulation when web crawlers are blocked, and use multimodal AI parsing screenshots as a backup solution.

[0026] According to one aspect of the present invention, the dynamic file knowledge base node is further configured to: allow users to set initial reference weights for uploaded files, and dynamically adjust the weights in response to user instructions or AI suggestions during process execution; when the conclusions of different files are contradictory, the AI ​​model automatically adjusts the credibility weights and explains the basis for the weighting to the user.

[0027] According to one aspect of the present invention, the supply chain resilience analysis node is further configured such that the supply chain relationship graph is a weighted dynamic graph, wherein the weights represent the degree of interdependence between enterprises.

[0028] According to one aspect of the invention, the AI ​​model automatically labels nodes in the graph with one or more of the following tags: geographical risk, industry risk, financial risk, and legal and operational risk.

[0029] According to one aspect of the present invention, the system is configured to: continuously accumulate risk control process data, human intervention decisions and final conclusions during process execution, and use this data to train and optimize the AI ​​model to achieve self-evolution of the risk control strategy.

[0030] A smart risk analysis method based on browser plugins includes the following steps:

[0031] Process construction steps: Through the visual interface provided by the browser plugin, the analysis process defined by the user by dragging and connecting modular risk control nodes is received;

[0032] Process execution steps: Start the process engine to execute the analysis process, driven by the AI ​​model as the decision center;

[0033] Interactive control steps:

[0034] During process execution, monitor whether a preset interruption waiting node has been reached or whether a user's natural language instruction has been received; if so, pause the current process and wait for user interaction, which may include uploading files or providing supplementary information; parse the intent of the natural language instruction, and based on the parsed intent, control the process to jump to the target node to continue execution, or backtrack to the already executed node to re-execute;

[0035] Analysis and comparison steps: After a node is re-executed, the AI ​​model compares and analyzes the execution results of the node in different contexts and generates a comparison report;

[0036] Conclusion generation steps: After the analysis process is completed, the output results of all risk control nodes are integrated to generate structured risk analysis conclusions.

[0037] Advantages of this invention: Through the above technical solution, this invention successfully upgrades commercial factoring risk assessment from a labor-intensive operation relying on human experience to an AI-driven process. Its most significant advantage lies in achieving deep human-machine collaboration and dynamic system evolution: On the one hand, the system seamlessly integrates the professional judgment of human experts with the powerful data processing capabilities of AI through an "interruptible and interactive" mechanism, forming a "1+1>2" decision-making loop. This significantly improves assessment efficiency and accuracy while freeing risk control personnel from tedious and repetitive tasks, allowing them to focus on critical decisions. On the other hand, the system possesses a self-learning ability that becomes increasingly intelligent with use. Every process execution, manual intervention, and file upload becomes training data, continuously optimizing the AI ​​model's analysis strategy, enabling risk control insights to move from "surface scoring" to "deep prediction," ultimately constructing an intelligent risk control organism that can proactively adapt to complex business scenarios and continuously improve itself. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart illustrating the intelligent risk analysis research system and method based on a browser plugin as described in this invention. Detailed Implementation

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] like Figure 1 As shown, an intelligent risk analysis and research system based on a browser plugin is disclosed, the system comprising:

[0042] A visual workflow orchestration interface, loaded in a browser, is used to receive workflow orchestrations of modular risk control nodes by users through drag-and-drop and connection operations;

[0043] A process engine is used to execute the analysis process consisting of the risk control nodes;

[0044] The AI ​​model is coupled with the process engine and risk control nodes.

[0045] The system is configured as follows:

[0046] During process execution, in response to reaching a preset interruption waiting node or receiving a natural language command from the user, the current process is paused and the user is waiting for interaction, which may include uploading files or providing supplementary information.

[0047] The intent of the natural language instruction is parsed, and the process jumps to the target node in the flow to continue execution or backtracks to the previous node to re-execute, based on the intent.

[0048] After the node is re-executed, the AI ​​model compares and analyzes the execution results of the node in different contexts and generates a comparison report.

[0049] After the process is completed, the outputs of each node are integrated to generate a structured risk analysis conclusion.

[0050] The core of this invention lies in a visually orchestratable browser plugin system with an AI model as the decision-making center. It is not simply process automation, but an "AI risk control expert" with perception, decision-making, and evolutionary capabilities.

[0051] The system is described as follows:

[0052] 1. Plugin initialization and interface loading:

[0053] When the plugin is launched in the browser, a visual workflow orchestration and execution interface will be displayed as a sidebar on the right side of the browser window by default. Users can manually enlarge it to full screen for a better user experience.

[0054] 2. AI-driven intelligent processes:

[0055] Users drag and drop the required functional nodes from a pre-built risk control node library onto the canvas and define their execution order using connecting lines, thereby constructing a complete and customized risk control analysis workflow. However, the workflow is not fixed and has the following characteristics:

[0056] 2.1 Open Process Construction and AI Autonomy:

[0057] 2.1.1 AI-Assisted Orchestration Mode: After the user sets the core objectives, the AI ​​can automatically recommend and add key nodes based on its initial understanding of the target company. For example, when the target company is a high-tech company, the AI ​​will automatically add the "In-depth Intellectual Property Analysis" node.

[0058] 2.1.2 AI Fully Autonomous Mode: Users only need to input the company name, and the AI ​​will autonomously utilize its knowledge base and reasoning capabilities to dynamically construct and execute a complete analysis process, ultimately directly outputting risk control conclusions. In this mode, the AI's analysis path is not fixed but dynamically generated based on real-time analysis results.

[0059] 2.2 Deep Interactivity:

[0060] 2.2.1 During process execution, AI monitors and understands the context throughout. Supports:

[0061] (1) Interruption and waiting: AI can actively pause the process. When the process reaches a node that requires human intervention, the process engine will automatically pause and wait for input from risk control personnel, requesting the user to upload key documents (such as special audit reports).

[0062] (2) Command Interruption: Users can intervene at any time through natural language commands, such as "Skip the financial statement analysis and directly compare its litigation situation with that of the industry leader in the past three years." The AI ​​can understand the intent of the command and achieve seamless switching while preserving the current state.

[0063] 2.3 Dynamic Flexibility: Comparative Analysis of Process Backtracking and AI

[0064] The system supports arbitrary rollback and re-execution of processes. During process execution, risk control personnel and AI models form a closed-loop collaborative workflow. The core capability of AI lies in its ability to intelligently compare and analyze the execution results of the same node in different contexts. Based on real-time intermediate conclusions, AI can intelligently infer the analytical dimensions that should be strengthened subsequently and dynamically adjust the execution depth or path of subsequent nodes. The entire process forms a feedback loop that becomes smarter with use. By accumulating historical process data, the system continuously learns and optimizes risk control strategies, thereby achieving a leap from "process automation" to "intelligent analysis."

[0065] 2.3.1 Example: A user initially uploaded two industry reports (70% weighting and 30% weighting) to the "Document Knowledge Base" node. Later in the process, the AI ​​detected an anomaly in the user's accounts receivable turnover rate and suggested the user revert to the "Document Knowledge Base" node, add a new "Accounts Receivable Special Verification Report," and adjust its weighting. The AI ​​will generate a "Before and After Analysis Comparison Report," clearly indicating how the introduction of new evidence altered the assessment of the user's cash flow health and quantifying the impact of this change.

[0066] 2.4 Continuously improve existing conclusions and generate structured risk control conclusions.

[0067] After the process is completed, the system automatically integrates the output results of all nodes and generates a structured risk control report, which includes the scores of each stage, key findings, risk warnings, and the final conclusion and recommendation of "pass / fail / conditionally pass".

[0068] 3. Nodes where AI deeply empowers:

[0069] For the automated information acquisition and analysis nodes, the plugin automatically acquires and processes data from multiple channels through preset logic tailored to the risk control needs of factoring business. During this process, the deeply integrated AI model not only contributes to the final report generation but also plays a crucial role in data acquisition, cleaning, and cross-validation.

[0070] The following node will be used as an example to analyze in detail the implementation of its core functions:

[0071] 3.1 Enterprise Panoramic Insight Node

[0072] The design goal of this node is to reliably acquire multi-dimensional data of target companies from various enterprise information platforms (such as Qichacha and Tianyancha). This node employs an intelligent hierarchical strategy for data acquisition: prioritizing API and web crawler calls, enabling RPA simulation when blocked, and finally using multimodal AI-analyzed screenshots as a backup. AI ensures the success and quality of data acquisition throughout the entire process.

[0073] After acquiring enterprise information from multiple sources, the AI ​​will perform the following:

[0074] 3.1.1 Relationship Graph Mining: Automatically construct complex relationship networks among enterprises, shareholders, and senior executives, and use AI to identify potential risk structures such as circular shareholding and hidden actual controllers.

[0075] 3.1.2 Risk Signal Penetration: AI does not list information such as judicial proceedings, administrative penalties, and operational anomalies in isolation, but rather conducts spatiotemporal correlation analysis. For example: "In the past year, the number of cases in which the company was the defendant increased by 200% year-on-year, and most of them were concentrated in the supply chain field. This is highly correlated with its main business, indicating that there is a significant risk to the stability of its supply chain partnerships."

[0076] 3.2 Multidimensional Financial Statement Deep Diagnosis Node

[0077] AI deeply integrates the three reports to perform cross-period, cross-subject, and cross-industry collaborative analysis:

[0078] 3.2.1 Counterintuitive Reasoning: When a low debt-to-equity ratio is detected, the AI ​​will not directly conclude that the company is "safe." Instead, it will automatically correlate the "cash flow from investing activities" in the "cash flow statement" with the "other income" in the "profit and loss statement." If the AI ​​finds that the company is selling off assets in large quantities or receiving unconventional subsidies, it will point out: "The low debt ratio stems from the disposal of non-core assets, its core business's ability to generate cash is actually worrying, and its short-term solvency is an illusion."

[0079] 3.2.2 Trend Prediction and Stress Testing: Based on historical data, AI predicts the profitability and cash flow situation for the next year and can simulate the company's risk resistance under macroeconomic downturn (such as a 10% decrease in revenue).

[0080] 3.3 Dynamic File Knowledge Base Node

[0081] This node allows risk control personnel to upload any information files related to the risks of the companies they are reviewing, forming a new form of "knowledge base" that feeds back into the AI ​​model.

[0082] 3.3.1 Multimodal Understanding and Weight Learning: Users can upload any document, such as brokerage research reports, internal reviews, and press releases. The AI ​​can not only read text but also understand tables and charts. Users can set initial reference weights for the documents and dynamically adjust the weight values ​​during execution based on specific circumstances.

[0083] 3.3.2 Conflict of Evidence Resolution: When different documents reach contradictory conclusions (e.g., a research report is optimistic but internal reviewers are pessimistic), the AI ​​will analyze the source, timeliness, and logical chain of the evidence, automatically adjust the credibility weight, and explain the basis for its weighting to the user.

[0084] 3.3.3 Knowledge Accumulation: All uploaded files and their final conclusions will be vectorized and stored in the AI's knowledge base to enhance the knowledge background for all future analyses.

[0085] 3.4 Intelligent Analysis Nodes in Judicial Proceedings

[0086] In addition to performing routine case count statistics, this node also uses AI to perform the following tasks:

[0087] 3.4.1 In-depth analysis of the case: Using NLP technology, interpret the original text of the judgment to identify the business logic and risk patterns behind the case.

[0088] 3.4.2 Impact Assessment: Quantitatively analyze the potential impact of pending litigation on the target company's future cash flow, goodwill, and financing capabilities, and provide a probability assessment.

[0089] 3.5 Supply Chain Resilience Analysis Node

[0090] This node aims to fundamentally change the static and one-sided understanding of supply chain risk in traditional risk control. It no longer simply lists suppliers and customers, but instead uses AI to proactively construct a "dynamic digital twin of the target company's supply chain" and, by simulating external shocks, proactively assesses the robustness and vulnerability of its supply chain network. This is a predictive analytics node entirely driven by AI and data.

[0091] 3.5.1 AI-powered automatic recognition and map construction:

[0092] The first step is to integrate multi-source data: AI automatically captures, identifies and verifies the target company's core upstream suppliers (especially exclusive or key suppliers) and downstream core customers from enterprise panoramic information, bidding announcements, news and public opinion, industry research reports, and even the company's official website and product descriptions.

[0093] Then, a dynamic relationship network is constructed: instead of building a static list, the AI ​​generates a weighted, dynamic supply chain relationship graph. In this graph, nodes represent companies, lines represent relationships, and the AI ​​assigns a dependency weight to each relationship (for example, if purchases from supplier A account for more than 60% of the total purchase amount, the dependency weight is marked as "critical").

[0094] 3.5.2 Multi-dimensional risk database injection:

[0095] AI seamlessly integrates external macroeconomic risk data into the aforementioned graph. It automatically labels each supplier and customer node in the graph with multi-dimensional risk tags, for example:

[0096] (1) Geographical risks: Whether the company’s registered location or main production area is located in a geopolitical hotspot or an area prone to natural disasters.

[0097] (2) Industry risks: Whether the industry in which the company operates is in a period of policy regulation or whether it faces technological obsolescence.

[0098] (3) Financial risk: The company’s own publicly disclosed financial health status (obtained by linking the “financial statement analysis node”).

[0099] (4) Legal and operational risks: Does the company have serious legal disputes or abnormal business records?

[0100] 3.5.3 Intelligent Simulation and Resilience Modeling:

[0101] Based on the constructed supply chain map and risk labels, AI will automatically run various "stress test" scenarios to simulate the transmission path and impact of black swan events on target companies.

[0102] 3.5.4 The results of each simulation, whether they actually occur or are merely hypothetical, are recorded and learned by the AI. When a similar event occurs in the real world, the AI ​​will automatically review its historical simulation predictions, compare them with the actual results, and thus optimize its simulation model parameters to make its next prediction more accurate.

[0103] This node elevates risk control from "post-event statistics" to "pre-event prediction," and from "single-point observation" to "system simulation." It empowers risk control personnel with an unprecedented capability: to anticipate potential supply chain crises within a given timeframe and visualize their quantifiable consequences, enabling proactive planning and truly achieving proactive risk management. This fully demonstrates the core value of AI as a decision-making brain.

[0104] Advantages of this invention: Through the above technical solution, this embodiment successfully transforms the traditional static risk control pipeline into a dynamic, intelligent, and evolvable decision support system. Its core effect lies in achieving a leapfrog improvement in efficiency, depth, and flexibility: In terms of efficiency, the system, through automated execution and AI-driven processes, liberates risk control personnel from tedious multi-platform switching and data transfer, greatly shortening the analysis cycle and significantly reducing human error; in terms of analytical depth, the AI ​​model's multi-dimensional correlation analysis, counterintuitive reasoning, and predictive simulation capabilities can penetrate surface data to discover hidden risks and complex patterns that are difficult for humans to detect, elevating risk assessment from "post-event statistics" to "pre-event prediction"; in terms of flexibility, the unique interactive process design (interruption, jump, and backtracking) achieves seamless integration of human and machine intelligence, enabling the risk control process to dynamically adjust with business scenarios rather than being rigidly executed. Ultimately, the system continuously learns from each interaction and decision, forming a virtuous cycle that becomes increasingly accurate with use, building a continuously evolving risk immunity capability for enterprises.

[0105] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A smart risk analysis and research system based on a browser plugin, characterized in that, The system includes: The workflow orchestration interface is used to receive users' visual orchestration operations on multiple risk control nodes to form a risk analysis workflow; The process execution engine is used to run the risk analysis process. Artificial Intelligence (AI) module; The process execution engine works in conjunction with the AI ​​decision-making module, enabling the execution path of the risk analysis process to be dynamically adjusted by the AI ​​decision-making module based on the intermediate results of the analysis process.

2. The intelligent risk analysis and research system based on browser plugins according to claim 1, characterized in that, The system is in the form of a browser plugin. The process arrangement interface is presented in the browser and can switch between sidebar and full-screen mode in response to user operations.

3. The intelligent risk analysis and research system based on browser plugins according to claim 1, characterized in that, The process execution engine supports at least two process construction modes: AI-assisted orchestration mode: In response to the analysis goals set by the user, the AI ​​model recommends and automatically adds relevant risk control nodes to the orchestration interface; AI Fully Autonomous Mode: In response to the company name entered by the user, the AI ​​model autonomously constructs and executes a complete analysis process.

4. The intelligent risk analysis and research system based on browser plugins according to claim 1, characterized in that, The risk control node includes at least one of the following: The enterprise panoramic insight node is configured to acquire enterprise information and build a relational graph through one or more of the following methods: API interface, web crawler, RPA robotic process automation, or multimodal AI analysis. The multi-dimensional financial statement deep diagnostic node is configured to perform cross-period and cross-account linkage analysis on the company's balance sheet, income statement, and cash flow statement, and to perform counterintuitive reasoning and stress testing. The dynamic file knowledge base node is configured to allow users to upload and analyze related files, and the AI ​​decision-making module performs multimodal understanding, weight learning, and evidence conflict resolution on the file content. The intelligent analysis node for judicial litigation is configured to use natural language processing technology to analyze judgment documents and conduct in-depth analysis of the cause of action and assessment of the litigation impact. The supply chain resilience analysis node is configured to construct a supply chain relationship graph of the target enterprise by the AI ​​decision-making module, inject multi-dimensional risk data, and run stress test scenarios simulating external shocks.

5. The intelligent risk analysis and research system based on browser plugins according to claim 4, characterized in that, The enterprise panoramic insight node is also configured to: acquire data using an intelligent hierarchical strategy, prioritize calling API interfaces, enable web crawlers when API interface calls fail, enable RPA simulation when web crawlers are blocked, and use multimodal AI parsing screenshots as a backup solution.

6. The intelligent risk analysis and research system based on browser plugins according to claim 4, characterized in that, The dynamic file knowledge base node is also configured to: allow users to set initial reference weights for uploaded files, and dynamically adjust the weights in response to user instructions or AI suggestions during process execution; when the conclusions of different files are contradictory, the AI ​​decision module automatically adjusts the credibility weights and explains the basis for the weighing to the user.

7. The intelligent risk analysis and research system based on browser plugins according to claim 4, characterized in that, The supply chain resilience analysis node is further configured such that the supply chain relationship graph is a weighted dynamic graph, where the weights represent the degree of interdependence between enterprises.

8. The intelligent risk analysis and research system based on browser plugins according to claim 7, characterized in that, The AI ​​decision-making module automatically labels nodes in the graph with one or more of the following tags: geographical risk, industry risk, financial risk, and legal and operational risk.

9. The intelligent risk analysis and research system based on browser plugins according to claim 1, characterized in that, The system is configured to continuously accumulate risk control process data, human intervention decisions, and final conclusions during process execution, and use this data to train and optimize the AI ​​decision-making module to achieve self-evolution of risk control strategies.

10. A smart risk analysis method based on browser plugins, characterized in that, The intelligent risk analysis and research system based on browser plugins as described in any one of claims 1 to 9 includes the following steps: Process receiving steps: Used to receive users' visual orchestration operations on multiple risk control nodes to form a risk analysis process; Process execution steps: Used to run the risk analysis process; AI interaction steps.

Images