Data asset management and monitoring method based on artificial intelligence technology
Through artificial intelligence-based data asset management and monitoring methods, the problem that traditional asset management methods cannot process real-time data and provide accurate risk warnings is solved, and efficient and intelligent asset management and decision-making is achieved, ensuring the accuracy and fairness of decision-making , and find a balance between privacy protection and model performance.
Patent Information
- Application Number
- CN202411354099.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-27
- Publication Date
- 2025-01-07
AI Technical Summary
Traditional asset management methods rely on manual judgment and fixed rules, and cannot process a large amount of real-time data, resulting in inefficient decision-making and difficulty in providing comprehensive and accurate asset status monitoring and risk warning functions.
Using data asset management and monitoring methods based on artificial intelligence technology, by collecting asset data from multiple data sources, performing pre-processing and machine learning analysis, generating asset management models, monitoring asset status in real time, making risk predictions, generating decision-making suggestions, and Dynamically adjust the model through technical means such as incremental learning and fairness assessment to ensure the efficiency and impartiality of decision-making.
It realizes efficient, reliable and intelligent data asset management, which can dynamically respond to market changes, improve the accuracy and fairness of decisions, and find a balance between privacy protection and model performance.
Smart Images
Figure CN119272006A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data asset management and monitoring technology, and in particular to a data asset management and monitoring method based on artificial intelligence technology. Background Art
[0002] With the development of information technology, data is increasingly used in various industries, especially in the field of asset management, where data has become an important resource for promoting decision-making and optimizing management. Traditional asset management methods rely on manual judgment and fixed rules, cannot handle large amounts of real-time data, and are prone to lag and subjectivity problems, resulting in inefficient decision-making. At the same time, with the surge in data volume, traditional asset management systems are unable to effectively process and analyze this data, and are unable to provide comprehensive and accurate asset status monitoring and risk warning functions. Summary of the invention
[0003] In view of this, the object of the present invention is to provide a data asset management and monitoring method based on artificial intelligence technology, so as to provide efficient, safe and reliable management and monitoring of data assets.
[0004] The present invention discloses a data asset management and monitoring method based on artificial intelligence technology, which comprises the following steps:
[0005] S1. Collect asset data from multiple data sources, where the asset data includes real-time asset data and historical asset data;
[0006] S2. performing preprocessing operations on the collected asset data, wherein the preprocessing operations include data cleaning and standardization;
[0007] S3. Analyze the data after the preprocessing operation through machine learning algorithms to generate an asset management model;
[0008] S4. Based on the asset management model, monitor the status of the assets in real time, perform risk prediction, and obtain prediction results;
[0009] S5. Generate asset management decision suggestions based on the prediction results and perform corresponding asset management operations.
[0010] Further, the machine learning algorithm includes a support vector machine algorithm;
[0011] In the process of generating the asset management model, determining the asset management objectives; wherein the asset management objectives at least include asset return objectives, asset risk objectives, and asset liquidity objectives;
[0012] Perform multi-objective optimization operations on asset management goals and assign corresponding weights to each goal;
[0013] Extracting nonlinear features of asset data, performing weighted combination of multiple kernel functions based on the nonlinear features, and automatically adjusting weights of different kernel functions to obtain an adaptive kernel function;
[0014] A support vector machine model is constructed based on the adaptive kernel function, and the coefficients of the support vector machine model and the weights of the kernel function are regularized, and the support vector machine model is used as an asset management model.
[0015] Furthermore, the step S3 also includes dynamically updating the asset management model through an incremental learning algorithm, specifically including:
[0016] Calculate new support vectors based on the collected new asset data, update the support vector set of the model based on the new support vectors, and adjust the Lagrange multiplier;
[0017] Recalculate model parameters based on the incremental features of new asset data and update model parameters;
[0018] Extract the nonlinear features of the new asset data and optimize the weights of the kernel function based on the nonlinear features of the new asset data.
[0019] Furthermore, step S3 also includes performing an algorithm fairness evaluation operation during the training process of the asset management model.
[0020] Furthermore, the algorithm fairness evaluation operation specifically includes:
[0021] Define fairness indicators, including equal opportunity indicators, equilibrium error rates, and group fairness indicators;
[0022] Establish a multi-level fair assessment mechanism, including asset category level, time dimension and customer group level, with each level being a group;
[0023] Construct a causal relationship diagram based on the causal relationships between different groups;
[0024] During each iteration of the asset management model, the fairness value of the model in each group is dynamically evaluated based on the defined fairness indicators and causal relationship diagram, and the real-time evaluation results are fed back to the model training process, and the model training parameters are dynamically adjusted based on the evaluation results.
[0025] Furthermore, the method further comprises:
[0026] Conduct risk assessment on asset data of each group and classify risk levels based on risk assessment results;
[0027] According to the risk level of each group of asset data, a differential privacy budget is initially allocated to each group of asset data;
[0028] During each iteration of model training, data characteristics and model learning progress are monitored in real time to determine the need for privacy protection in the current training phase.
[0029] Adaptively adjust the intensity and method of noise injection according to real-time monitoring results and risk levels of data;
[0030] At the end of each training iteration, real-time feedback and analysis are performed based on the model's performance and the current privacy protection effect to evaluate whether the current privacy protection reaches the preset value.
[0031] If the preset value is not reached, the privacy budget allocation and noise injection strategy are dynamically adjusted.
[0032] Furthermore, the step S2 includes performing a data preprocessing operation through a data enhancement algorithm.
[0033] Furthermore, the data preprocessing operation performed by the data enhancement algorithm specifically includes:
[0034] Generate new asset price sequences as new training data samples based on existing historical asset data by randomly increasing the amplitude of price fluctuations or changing the cycle of market fluctuations;
[0035] Simulation-based data augmentation technology simulates extreme market conditions and generates asset data under extreme market conditions as new training data samples.
[0036] Furthermore, before executing the corresponding asset management operation in step S5, a manual intervention review mechanism is also included. When the generated asset management decision suggestion does not meet the expectation, it is reviewed and adjusted by experts.
[0037] Compared with the prior art, the present invention has the following beneficial effects:
[0038] The present invention manages and monitors data assets based on artificial intelligence technology, builds and trains models through support vector machine algorithms and asset data obtained from different sources, and generates an asset management model. Based on the asset management model, asset data can be analyzed and effective asset management suggestions can be output, thereby achieving efficient, reliable and intelligent data asset management. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:
[0040] Figure 1 A flowchart of a data asset management and monitoring method based on artificial intelligence technology disclosed in an embodiment of the present invention. DETAILED DESCRIPTION
[0041] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments.
[0042] Embodiment 1
[0043] See also Figure 1 , Figure 1 1 is a flow chart of a data asset management and monitoring method based on artificial intelligence technology disclosed in an embodiment of the present invention, the method comprising the following steps:
[0044] S1. Collect asset data from multiple data sources, where the asset data includes real-time asset data and historical asset data.
[0045] Specifically, multiple data sources refer to different sources for obtaining asset data, including different systems, platforms or data providers, such as company financial reports, banks and financial institutions, macroeconomic data sources, etc. Asset data includes but is not limited to stock prices, bond data, money market data, futures and options data, cryptocurrency data in real-time asset data, historical price data, asset transaction records, historical return and risk data, financial data history in historical asset data, asset characteristic data such as asset categories, asset ratings, liquidity data, industry data, macroeconomic and market environment data such as economic indicators, market sentiment data, policy change data.
[0046] S2. Perform preprocessing operations on the collected asset data, wherein the preprocessing operations include data cleaning and standardization.
[0047] Specifically, in the embodiment of the present invention, data cleaning of asset data includes but is not limited to deleting missing values, interpolation filling, mean or median filling, outlier processing, and data duplication processing. After data cleaning, standardization processing is performed to ensure that all data features are within a uniform scale range to avoid certain features from having unnecessary effects in model training.
[0048] S3. Analyze the data after preprocessing operations through machine learning algorithms to generate an asset management model.
[0049] S4. Based on the asset management model, the status of the assets is monitored in real time, and risk prediction is performed to obtain prediction results.
[0050] S5. Generate asset management decision suggestions based on the prediction results and perform corresponding asset management operations.
[0051] Furthermore, the machine learning algorithm includes a support vector machine algorithm, specifically including:
[0052] In the process of generating the asset management model, determining the asset management objectives; wherein the asset management objectives at least include asset return objectives, asset risk objectives, and asset liquidity objectives;
[0053] Perform multi-objective optimization operations on asset management goals and assign corresponding weights to each goal;
[0054] Extracting nonlinear features of asset data, performing weighted combination of multiple kernel functions based on the nonlinear features, and automatically adjusting weights of different kernel functions to obtain an adaptive kernel function;
[0055] A support vector machine model is constructed based on the adaptive kernel function, and the coefficients of the support vector machine model and the weights of the kernel function are regularized, and the support vector machine model is used as an asset management model.
[0056] Furthermore, the asset management model is specifically:
[0057]
[0058] The constraints are:
[0059]
[0060] Among them, α k is the Lagrange multiplier of the support vector machine, k represents different asset management objectives, K is the number of objectives, and w k is the target weight; m is the weight of the adaptive kernel function, m is the index of the kernel function, M is the number of kernel functions used in the model; C is the regularization parameter; is the loss function, which is used to measure the prediction result f(x j ) and the true label y j The gap between m (x j ,x i ) is the kernel function; γ is the regularization parameter of the kernel function weight; N is the number of samples of training data, x j is the input feature vector of the jth training sample, x i is the i-th support vector.
[0061] In an embodiment of the present invention, an asset management model is constructed by combining multi-objective optimization, adaptive kernel function adjustment, and regularization processing, so that the model can achieve a dynamic balance between asset management objectives such as returns, risks, and liquidity, and adapt to different market demands and investment strategies. Among them, the kernel function is adjusted based on the extraction of nonlinear features, so that the model can effectively cope with complex market dynamics and nonlinear relationships between assets, thereby improving prediction accuracy and decision-making capabilities. In addition, by further combining regularization processing, the model has better robustness and robustness in responding to market changes, reducing the risk of overfitting.
[0062] Furthermore, step S3 also includes dynamically updating the asset management model through an incremental learning algorithm, specifically including:
[0063] Calculate new support vectors based on the collected new asset data, update the support vector set of the model based on the new support vectors, and adjust the Lagrange multiplier;
[0064] Recalculate model parameters based on the incremental features of new asset data and update model parameters;
[0065] Extract the nonlinear features of the new asset data and optimize the weights of the kernel function based on the nonlinear features of the new asset data.
[0066] The core feature of the incremental learning algorithm is that the model can be updated when new data is obtained without having to retrain the entire model. In asset management, the market environment and asset performance may change significantly over time. Through incremental learning, the model can dynamically adapt to these changes, thereby improving the real-time and accuracy of decision-making. When the market fluctuates, asset prices change, or new asset data is introduced, the model can be updated in a timely manner to reflect the latest market information, helping asset managers make more accurate investment decisions in a dynamic environment. In addition, through incremental learning, the computational cost of the model can be reduced, especially when dealing with large-scale asset data or real-time asset data, allowing the model to be updated more efficiently and kept up to date,
[0067] Furthermore, step S3 also includes performing an algorithm fairness evaluation operation during the training process of the asset management model.
[0068] Furthermore, the algorithm fairness evaluation operation specifically includes:
[0069] Define fairness indicators, including equal opportunity indicators, equilibrium error rates, and group fairness indicators;
[0070] Establish a multi-level fair assessment mechanism, including asset category level, time dimension and customer group level, with each level being a group;
[0071] Construct a causal relationship diagram based on the causal relationships between different groups;
[0072] During each iteration of the asset management model, the fairness value of the model in each group is dynamically evaluated based on the defined fairness indicators and causal relationship diagram, and the real-time evaluation results are fed back to the model training process, and the model training parameters are dynamically adjusted based on the evaluation results.
[0073] Specifically, the equal opportunity indicator is used to ensure that the model has the same chance to make the right decision in different asset classes or customer groups. For example, the model's return forecasts for high-net-worth customers and ordinary customers should have the same accuracy. The balanced error rate is used to evaluate the error rate of the model in different groups to ensure that the error rates in different asset classes or customer groups are not too different. The group fairness indicator is used to measure the performance of the model in various customer groups to ensure that no group is systematically ignored or mistreated.
[0074] When establishing a multi-level fairness assessment mechanism, by dividing the data into multiple levels, such as asset category level (stocks, bonds, funds, etc.), time dimension (historical data, real-time data, etc.), customer group level (high net worth customers, ordinary customers, etc.), these different levels represent different assessment groups, so that the model conducts fairness assessments at each level. By conducting fairness assessments on multiple dimensions, it is ensured that the model's decisions in different asset categories, time periods, and customer groups are consistent and fair, avoiding the model's deviation in a specific dimension.
[0075] In an embodiment of the present invention, a causal relationship graph is also constructed based on the potential causal relationships between different groups. The causal relationship graph is used to reveal the implicit dependencies that may exist in the model decision-making process, for example, the causal relationship between the asset categories of certain customer groups and their investment returns. Through the causal relationship graph, the model can have a deeper understanding of the causal relationships in the data and avoid unfair decisions caused by ignoring certain key causal factors. For example, if the investment decisions of high-net-worth customers are unfairly manipulated in the model, the causal relationship graph can help discover this potential problem.
[0076] In the embodiment of the present invention, by performing an algorithm fairness evaluation operation during the training process of the asset management model, potential biases in the model can be effectively identified and corrected, ensuring that the decision-making process of the model in different asset categories, time dimensions, and customer groups is fair, reasonable, and consistent. This operation enables the model to not only have efficient performance, but also ensure fairness in the asset management process, improve the transparency and credibility of the model, and better meet the needs of practical applications.
[0077] Furthermore, the method further comprises:
[0078] Conduct risk assessment on asset data of each group and classify risk levels based on risk assessment results;
[0079] According to the risk level of each group of asset data, a differential privacy budget is initially allocated to each group of asset data;
[0080] During each iteration of model training, data characteristics and model learning progress are monitored in real time to determine the need for privacy protection in the current training phase.
[0081] Adaptively adjust the intensity and method of noise injection according to real-time monitoring results and risk levels of data;
[0082] At the end of each training iteration, real-time feedback and analysis are performed based on the model's performance and the current privacy protection effect to evaluate whether the current privacy protection reaches the preset value.
[0083] If the preset value is not reached, the privacy budget allocation and noise injection strategy are dynamically adjusted.
[0084] Specifically, risk assessment is conducted on asset data of each group based on different asset categories (such as stocks, bonds, funds, etc.), customer groups (such as high net worth customers, ordinary customers, etc.), and market environments (such as volatility, market uncertainty, etc.). Risk assessment includes but is not limited to analysis of factors such as asset volatility, historical performance, and market relevance. Through risk assessment, different asset data are divided into different risk levels (such as high risk, medium risk, and low risk), providing a basis for subsequent privacy budget allocation.
[0085] During each iteration of model training, data characteristics (such as data distribution, volatility, etc.) and the model's learning process (such as convergence speed, training error, etc.) are monitored in real time to evaluate the need for privacy protection in the current training stage. Through real-time monitoring, ensure that the model's needs for privacy protection at different training stages are responded to in a timely manner. For example, weaker privacy protection may be required in the early stages of training to improve the model's learning ability, while stronger privacy protection may be required in the later stages of training to prevent privacy leakage. Dynamically adjust the intensity and method of noise injection based on real-time monitoring results and the risk level of the data. For high-risk data, appropriately increase the intensity of noise injection to enhance privacy protection; for low-risk data, reduce noise injection to improve the accuracy of the model. By adaptively adjusting noise injection, more flexible privacy protection can be achieved without affecting model performance. For example, if the model is highly sensitive to certain data points, the privacy protection of these data can be strengthened. At the end of each training iteration, real-time feedback and analysis are performed based on the model's performance and the current privacy protection effect. By evaluating whether the current privacy protection reaches the preset value (such as whether the noise injection effect is appropriate), it is determined whether the strategy needs to be adjusted. Through the real-time feedback mechanism, it is ensured that the privacy protection effect is consistent with the expected goal, preventing the model from having insufficient or excessive privacy protection at certain stages.
[0086] If the privacy protection effect is found to be below the preset value in real-time feedback and analysis, the allocation of privacy budget and noise injection strategy will be adjusted dynamically. For example, the privacy budget can be reallocated, or the noise injection intensity of different data groups can be adjusted. The purpose of dynamic adjustment is to continuously optimize the privacy protection strategy during the training process to ensure that each data group can be trained at an appropriate privacy protection level.
[0087] Through the above operations, the flexibility, adaptability and dynamic optimization of privacy protection can be achieved during the training process of the asset management model. Not only does it protect the privacy of high-risk data, but it also enables the model to maintain efficient performance and good generalization capabilities during the training process. Through real-time monitoring and dynamic adjustment, the privacy protection measures of the model can be in a better state, thereby achieving a balance between data security and model performance.
[0088] Furthermore, step S2 includes performing a data preprocessing operation by a data enhancement algorithm.
[0089] Furthermore, the data preprocessing operation performed by the data enhancement algorithm specifically includes:
[0090] Generate new asset price sequences as new training data samples based on existing historical asset data by randomly increasing the amplitude of price fluctuations or changing the cycle of market fluctuations;
[0091] Simulation-based data augmentation technology simulates extreme market conditions and generates asset data under extreme market conditions as new training data samples.
[0092] Specifically, asset price sequences are extracted from existing historical asset data, and new asset price sequences are generated by introducing random fluctuations and changes in market cycles. This may include adding random fluctuations within a certain range to the original price data, or shortening or lengthening the price fluctuation cycle, thereby generating different price paths. By generating new asset price data in this way, the model can see price sequences with different fluctuation amplitudes and cycles, and adapt to different market conditions.
[0093] In addition, we use simulated data enhancement technology to generate asset data under extreme market conditions, simulating asset performance under abnormal market fluctuations, financial crises or other extreme situations. For example, we simulate the sharp drop in prices during a market crash, or the rapid price correction after a market bubble bursts.
[0094] In an embodiment of the present invention, new data samples generated by data enhancement are added to the training set of the model to expand the scale and diversity of the training data set. The new training samples include not only randomly fluctuating data, but also simulated data under extreme market conditions, so that the model can be trained in a wider range of scenarios and improve its performance in various market environments.
[0095] Furthermore, before executing the corresponding asset management operation in step S5, a manual intervention review mechanism is also included. When the generated asset management decision suggestion does not meet the expectation, it is reviewed and adjusted by experts.
[0096] In the embodiment of the present invention, asset management operations refer to specific execution behaviors in the asset management process, including but not limited to asset allocation, portfolio management, risk control, liquidity management, etc., such as adjusting the proportion of the investment portfolio, reallocating asset categories, performing risk hedging operations, dynamically adjusting liquidity management strategies, etc. Asset management decision suggestions are specific suggestions generated by the model based on market data and customer needs, which are used to guide asset management operations, including but not limited to asset allocation suggestions such as suggestions on the proportion of different asset categories (stocks, bonds, funds, etc.) in the investment portfolio; investment decision suggestions on how to buy or sell certain assets, recommended investment timing, holding period, etc.; risk management suggestions such as risk warnings for market fluctuations or specific assets, recommended risk hedging strategies or risk control measures; liquidity management suggestions such as suggestions on cash flow management, asset realization timing, etc., to ensure that the investment portfolio has sufficient liquidity.
[0097] Specifically, a review mechanism is established in the data asset management process to review the asset management decision recommendations generated by the model. When the model's decision recommendations do not meet the set expectations, the manual intervention process is automatically triggered and the recommendations are submitted to experts for review and adjustment. Specifically, by defining the expected standards related to the model recommendations, such as return targets, risk levels, liquidity requirements, etc., when the recommendations generated by the model deviate from these standards, they will be automatically marked and trigger manual intervention. For example, when the deviation between the asset allocation recommended by the model and the expected allocation exceeds a certain threshold, or when market conditions are abnormal, expert review is automatically triggered.
[0098] When the review mechanism is triggered, the expert team reviews the decision recommendations generated by the model to evaluate their rationality and feasibility. Based on the actual situation, the experts can adjust the recommendations or re-formulate the decision. After the experts complete the review and adjustment, the results are fed back to the model training process to help the model continuously optimize the quality of decision recommendations. Through feedback, the model can learn and improve, thereby reducing the number of discrepancies with expectations in future recommendations.
[0099] By establishing a review mechanism for manual intervention, a security layer can be added to the asset management system to ensure that the decision recommendations generated by the model are reviewed and optimized by experts at critical moments. This not only improves the reliability of the model, but also makes the asset management decision-making process more transparent and reasonable, and able to dynamically adapt to market changes and customer needs.
[0100] Finally, it should be noted that the data asset management and monitoring method based on artificial intelligence technology disclosed in the embodiment of the present invention is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A data asset management and monitoring method based on artificial intelligence technology, characterized in that: The method comprises the following steps: S1. Collect asset data from multiple data sources, where the asset data includes real-time asset data and historical asset data; S2. performing preprocessing operations on the collected asset data, wherein the preprocessing operations include data cleaning and standardization; S3. Analyze the data after the preprocessing operation through machine learning algorithms to generate an asset management model; S4. Based on the asset management model, monitor the status of the assets in real time, perform risk prediction, and obtain prediction results; S5. Generate asset management decision suggestions based on the prediction results and perform corresponding asset management operations.
2. The data asset management and monitoring method based on artificial intelligence technology according to claim 1 is characterized in that: The machine learning algorithm includes a support vector machine algorithm; In the process of generating the asset management model, determining the asset management objectives; wherein the asset management objectives at least include asset return objectives, asset risk objectives, and asset liquidity objectives; Perform multi-objective optimization operations on asset management goals and assign corresponding weights to each goal; Extracting nonlinear features of asset data, performing weighted combination of multiple kernel functions based on the nonlinear features, and automatically adjusting weights of different kernel functions to obtain an adaptive kernel function; A support vector machine model is constructed based on the adaptive kernel function, and the coefficients of the support vector machine model and the weights of the kernel function are regularized, and the support vector machine model is used as an asset management model.
3. The data asset management and monitoring method based on artificial intelligence technology according to claim 2 is characterized in that: The step S3 also includes dynamically updating the asset management model through an incremental learning algorithm, specifically including: Calculate new support vectors based on the collected new asset data, update the support vector set of the model based on the new support vectors, and adjust the Lagrange multiplier; Recalculate model parameters based on the incremental features of new asset data and update model parameters; Extract the nonlinear features of the new asset data and optimize the weights of the kernel function based on the nonlinear features of the new asset data.
4. The data asset management and monitoring method based on artificial intelligence technology according to claim 3 is characterized in that: The step S3 also includes performing an algorithm fairness evaluation operation during the training process of the asset management model.
5. The data asset management and monitoring method based on artificial intelligence technology according to claim 4 is characterized in that: The algorithm fairness evaluation operation specifically includes: Define fairness indicators, including equal opportunity indicators, equilibrium error rates, and group fairness indicators; Establish a multi-level fair assessment mechanism, including asset category level, time dimension and customer group level, with each level being a group.
6. The data asset management and monitoring method based on artificial intelligence technology according to any one of claim 5, characterized in that ,, the algorithm fairness evaluation operation specifically includes: Construct a causal relationship diagram based on the causal relationships between different groups; During each iteration of the asset management model, the fairness value of the model in each group is dynamically evaluated based on the defined fairness indicators and causal relationship diagram, and the real-time evaluation results are fed back to the model training process, and the model training parameters are dynamically adjusted based on the evaluation results.
7. The data asset management and monitoring method based on artificial intelligence technology according to any one of claim 6, characterized in that: The method further comprises: Conduct risk assessment on asset data of each group and classify risk levels based on risk assessment results; According to the risk level of each group of asset data, a differential privacy budget is initially allocated to each group of asset data; During each iteration of model training, data characteristics and model learning progress are monitored in real time to determine the need for privacy protection in the current training phase. Adaptively adjust the intensity and method of noise injection according to real-time monitoring results and risk levels of data; At the end of each training iteration, real-time feedback and analysis are performed based on the model's performance and the current privacy protection effect to evaluate whether the current privacy protection reaches the preset value. If the preset value is not reached, the privacy budget allocation and noise injection strategy are dynamically adjusted.
8. The data asset management and monitoring method based on artificial intelligence technology according to claim 1 is characterized in that: The step S2 includes performing a data preprocessing operation by a data enhancement algorithm.
9. The data asset management and monitoring method based on artificial intelligence technology according to claim 8 is characterized in that: The data preprocessing operation performed by the data enhancement algorithm specifically includes: Generate new asset price sequences as new training data samples based on existing historical asset data by randomly increasing the amplitude of price fluctuations or changing the cycle of market fluctuations; Simulation-based data augmentation technology simulates extreme market conditions and generates asset data under extreme market conditions as new training data samples.
10. The data asset management and monitoring method based on artificial intelligence technology according to claim 1 is characterized in that: The step S5 also includes establishing a manual intervention review mechanism before executing the corresponding asset management operation. When the generated asset management decision suggestion does not meet the expectations, it will be reviewed and adjusted by experts.