Digital asset management method and system based on intelligent perception

Abstract

The invention relates to the field of digital asset management, in particular to a digital asset management method and system based on intelligent perception. The method comprises the steps that feature data of all dimensions of digital assets are collected, the collected feature data of all the dimensions are cleaned and standardized, and a standardized data set is generated; calculating a feature entropy based on the standardized data set; constructing a feature mapping matrix based on the feature entropy; calculating a comprehensive risk score of the assets based on the feature mapping matrix; based on the comprehensive risk score, constructing an objective function, and balancing the income and the risk by optimizing the objective function; and solving the optimal value of the optimization objective function by using a Bayesian update-based adaptive weight adjustment algorithm. The problems that due to the fact that static preset rules are adopted in asset classification, classification results are not accurate enough, an income and risk balance mechanism is lacked, asset allocation strategies cannot be adjusted in a self-adaptive mode according to market fluctuation, allocation decisions are lagged, and flexibility is lacked are solved.

Description

Technical Field

[0001] The present invention relates to the field of digital asset management, and in particular to a digital asset management method and system based on intelligent perception. Background Art

[0002] The digital asset market is characterized by high volatility and rapid changes, and different types of digital assets vary significantly in market performance, risk characteristics, liquidity, and holder reputation. This dynamic and complex market environment places higher demands on the real-time and intelligent management of digital assets, making it difficult for traditional static and manual management methods to cope with it.

[0003] In the application scenarios of digital assets, investors and managers need to be able to grasp the dynamic changes of assets in real time, make timely assessments of their potential risks, and adjust asset allocation accordingly to ensure maximum returns and minimum risks. For example, when the market fluctuates greatly, the management system needs to automatically identify high-risk assets and reduce their allocation ratio, while when the market tends to stabilize, assets with high return potential should be allocated first. This demand requires digital asset management to not only collect basic information about assets, but also to be able to perform adaptive analysis and optimize allocation based on changes in real-time data. With the rapid changes in the digital asset market and the diversification of investor needs, the management system needs to have the ability to quickly process massive, multi-dimensional data, including information such as asset type, market volatility, trading volume, and holder reputation. At the same time, the management system also needs to have certain intelligent analysis capabilities to automatically identify and evaluate the risk factors of assets, especially in market conditions with high uncertainty and volatility, to provide investors with robust risk control.

[0004] However, most existing digital asset management methods use static preset rules for asset classification, which leads to inaccurate classification results and the inability to effectively distinguish subtle differences in assets. The lack of a comprehensive risk quantification model leads to one-sided evaluation results and the inability to effectively identify potentially high-risk assets. At the same time, there is a lack of a balance mechanism between returns and risks, and the asset allocation strategy cannot be adaptively adjusted according to market fluctuations. It is impossible to achieve a balance between maximizing returns and controlling risks. The optimization process is difficult to make adaptive adjustments to market changes, and configuration decisions are delayed and lack flexibility, which affects the compliance and operational safety of the configuration. Summary of the invention

[0005] The present invention provides a digital asset management method and system based on intelligent perception to solve the problems that the use of static preset rules in asset classification will lead to inaccurate classification results, cannot effectively distinguish subtle differences in assets, lack of a comprehensive risk quantification model, resulting in one-sided evaluation results, cannot effectively identify potential high-risk assets, lack of a balance mechanism between return and risk, cannot adaptively adjust asset allocation strategies according to market fluctuations, cannot achieve a balance between maximizing returns and controlling risks, the optimization process is difficult to adaptively adjust according to market changes, configuration decisions are delayed, lack flexibility, and affect the compliance of configuration and operational safety.

[0006] The digital asset management method and system based on intelligent perception of the present invention specifically include the following technical solutions:

[0007] The digital asset management method based on intelligent perception includes the following steps:

[0008] S1. Collect feature data of each dimension of digital assets, clean and standardize the collected feature data of each dimension to generate a standardized data set; calculate feature entropy based on the standardized data set; construct a feature mapping matrix based on the feature entropy; calculate the comprehensive risk score of the asset based on the feature mapping matrix;

[0009] S2. Based on the comprehensive risk score, construct an objective function and balance the return and risk by optimizing the objective function; use the adaptive weight adjustment algorithm based on Bayesian update to solve the optimal value of the optimization objective function, maximize the return of digital assets and balance risk control.

[0010] Preferably, the S1 specifically includes:

[0011] Each dimension in the standardized data set is a feature, and the contribution of each feature in the classification is obtained by calculating the feature entropy. The specific calculation formula of the feature entropy is as follows:

[0012]

[0013] Among them, feature entropy H(X) represents the importance of each feature to the classification decision, which is used to measure the uncertainty of the feature set X and the weight influence of each feature in the classification process; w i is the weight of the i-th feature; x i is the value of the ith feature; n is the number of features; δ is a smoothing constant used to avoid zero values ​​when calculating logarithms; α is a weight adjustment factor; and μ is the feature mean.

[0014] Preferably, the S1 specifically includes:

[0015] By combining the risk factors of market volatility, asset liquidity and holder reputation, a comprehensive risk score is generated. The specific formula is as follows:

[0016]

[0017] Among them, S R represents the comprehensive risk score of the asset; m is the number of risk factors; is the weight of the kth risk factor; M v Indicates the degree of market volatility; β k is the adjustment coefficient, which is used to control the sensitivity of market fluctuations to the kth risk factor; C s Indicates the credit score of the holder; L i It is the liquidity of assets.

[0018] Preferably, the S2 specifically includes:

[0019] The configuration weight of each digital asset is dynamically adjusted based on the balance between returns and risks, and the maximization of the comprehensive rate of return of the asset is taken as the optimization goal. The objective function is constructed and the comprehensive risk score is used as a constraint in the optimization process. The specific formula of the objective function is as follows:

[0020]

[0021] Where I represents the objective function; A is the configuration weight of the u-th type of digital assets; u represents the current value of the u-th digital asset; γ is the market volatility sensitivity; R y Indicates the current rate of return; is the risk adjustment parameter; N is the total number of asset classes in the current asset allocation.

[0022] Preferably, the S2 specifically includes:

[0023] The adaptive weight adjustment algorithm based on Bayesian updating updates the distribution of configuration weights according to the latest observation data. The optimization process of the objective function is based on the gradient descent method. By gradually adjusting the configuration weights, the optimal configuration that maximizes the objective function is found.

[0024] Preferably, the S2 specifically includes:

[0025] The process of solving the optimization objective function includes:

[0026] Step 1: Calculate the Bayesian posterior probability; Calculate the posterior probability of each configuration weight based on the latest observation data and risk score;

[0027] Step 2: Initialize the weight distribution; use the mean or maximum posterior probability of the Bayesian posterior probability as the initial configuration, and initialize the weights of each configuration to the optimal configuration under the current conditions;

[0028] Step 3: Optimize the value of the objective function by gradient descent. After initializing the weight distribution, gradually adjust each configuration weight by gradient descent to find the configuration that maximizes the objective function.

[0029] Step 4: Iterative update and dynamic adaptation; recalculate the Bayesian posterior probability after each adjustment of the configuration weight, and dynamically adjust the configuration weight through iteration to maximize the objective function;

[0030] Step 5: During the gradient update process, each configuration weight satisfies the non-negativity and sum constraints.

[0031] Preferably, the S2 specifically includes:

[0032] Bayesian Posterior Probability The specific calculation formula is:

[0033]

[0034] in, Indicates that in the current weight configuration The probability of observing the latest observation data D under represents the prior probability; P(D) represents the overall probability of observing the latest observation data D; D u is the observed data related to the u-th asset; ∈ is a smoothing constant, which is used to prevent invalid results when the data value is zero; Give risk score to the u-th asset class.

[0035] The digital asset management system based on intelligent perception includes the following parts:

[0036] Data acquisition module, data preprocessing module, feature extraction module, feature mapping matrix construction module, risk assessment module, optimization configuration module, asset execution module;

[0037] The data collection module is responsible for real-time collection of various characteristic data of digital assets. Through the parallel data channel structure, various characteristic data are monitored in real time. The data collection module directly transmits the collected various characteristic data to the data preprocessing module.

[0038] The data preprocessing module cleans and standardizes various feature data to generate a standardized data set; each dimension in the standardized data set is a feature; the data preprocessing module passes the standardized data set to the feature extraction module;

[0039] The feature extraction module uses a standardized data set to calculate the feature entropy for each feature; the feature extraction module transfers the feature entropy of each feature to the feature mapping matrix construction module;

[0040] The feature mapping matrix construction module generates a feature mapping matrix according to the feature entropy; the feature mapping matrix construction module transmits the constructed feature mapping matrix to the risk assessment module;

[0041] The risk assessment module calculates the comprehensive risk score of each type of asset based on the feature mapping matrix; the risk assessment module passes the generated comprehensive risk score to the optimization configuration module;

[0042] The optimization configuration module performs asset configuration optimization based on the comprehensive risk score, updates the configuration weights according to the latest market feedback through the adaptive weight adjustment algorithm based on Bayesian update, and generates configuration instructions; the optimization configuration module transmits the optimal configuration weights to the asset execution module;

[0043] The asset execution module executes the asset allocation strategy according to the optimal configuration weight; the asset execution module receives the configuration instructions and feeds the results back to the data acquisition module after execution, forming a closed-loop system.

[0044] The beneficial effects of the technical solution of the present invention are:

[0045] 1. Comprehensively collect and monitor the core information of digital assets in real time, covering multi-dimensional feature data such as asset type, volatility, trading volume, and holder reputation; evaluate the weight of each feature in the classification by calculating feature entropy, and make weighted distribution according to the importance of each feature, thereby improving the accuracy of classification decisions; the introduction of feature entropy enables the digital asset management system to achieve more reasonable grouping in the process of automated classification by relying on the similarity and importance of features; the feature mapping matrix further describes the similarity relationship between features in high-dimensional space, thereby providing strong support for the automated classification of assets.

[0046] 2. By combining multiple risk factors such as market volatility, asset liquidity, and holder reputation, a comprehensive risk score is generated to quantify the risk level of various digital assets. The comprehensive risk score integrates multi-dimensional feature data into the evaluation process, making the digital asset management system more comprehensive and accurate in identifying potential high-risk assets. At the same time, the digital asset management system uses the comprehensive risk score as an important input factor for configuration optimization, helping managers make decisions based on risk levels, thereby effectively reducing the configuration ratio of high-risk assets.

[0047] 3. By optimizing the objective function, balancing returns and risks, and realizing dynamic optimization of asset allocation; the optimization objective function is reasonably designed and can adaptively adjust the asset allocation weights to ensure maximum returns even when the market fluctuates greatly; by introducing comprehensive risk scores and risk adjustment parameters, the digital asset management system strictly controls risks while maximizing returns, ensuring the stability and security of asset allocation; this balanced design enables the digital asset management system to find the best balance between returns and risks, thereby providing more reliable support for the management of digital assets.

[0048] 4. Adopt an adaptive weight adjustment algorithm based on Bayesian updating to update asset allocation weights in real time, so that the allocation strategy always conforms to the current market environment; Bayesian posterior probability provides a dynamic prior information, so that the optimization process has flexible adjustment capabilities to adapt to dynamic changes in the market; during the gradient descent process, the digital asset management system will gradually adjust the allocation weights to ensure that the allocation strategy can still efficiently converge to the optimal solution under different market conditions; through this adaptive closed-loop structure, the digital asset management system can respond to market changes in a timely manner to ensure the optimality and stability of the allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a structural diagram of the digital asset management system based on intelligent perception according to the present invention;

[0050] Figure 2 This is a flow chart of the digital asset management method based on intelligent perception described in the present invention. DETAILED DESCRIPTION

[0051] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0053] The specific scheme of the digital asset management method and system based on intelligent perception provided by the present invention is described in detail below with reference to the accompanying drawings.

[0054] Refer to the attached Figure 1 , which shows a structure diagram of a digital asset management system based on intelligent perception provided by an embodiment of the present invention, the system includes the following parts:

[0055] Data acquisition module, data preprocessing module, feature extraction module, feature mapping matrix construction module, risk assessment module, optimization configuration module, asset execution module;

[0056] The data collection module is responsible for real-time collection of multi-dimensional feature data of digital assets. Through a parallel data channel structure, various types of feature data are monitored in real time. The data collection module directly transmits the collected multi-dimensional feature data to the data preprocessing module.

[0057] The data preprocessing module cleans and standardizes the multi-dimensional feature data transmitted by the data acquisition module to generate a standardized data set for subsequent use; each attribute (or dimension) in the standardized data set is a feature; the data preprocessing module transfers the standardized data set to the feature extraction module;

[0058] The feature extraction module uses the standardized data set transmitted by the data preprocessing module to calculate the feature entropy for each feature to measure the importance of each feature in the classification process; the feature extraction module transmits the feature entropy of each feature to the feature mapping matrix construction module;

[0059] The feature mapping matrix construction module generates a feature mapping matrix in a multidimensional space according to the feature entropy transmitted by the feature extraction module to describe the similarity relationship between the features; the feature mapping matrix construction module transmits the constructed feature mapping matrix to the risk assessment module;

[0060] The risk assessment module calculates the comprehensive risk score of each type of asset based on the feature mapping matrix, combined with risk factors such as market volatility, asset liquidity and holder reputation, to quantify the risk level of each type of asset; the risk assessment module passes the generated comprehensive risk score to the optimization configuration module;

[0061] The optimization configuration module performs asset configuration optimization based on the comprehensive risk score transmitted by the risk assessment module. Through the adaptive weight adjustment algorithm based on Bayesian update, it updates the configuration weight according to the latest market feedback in each iteration and generates configuration instructions. The optimization configuration module transmits the optimal configuration weight to the asset execution module.

[0062] The asset execution module executes specific asset allocation strategies according to the optimal configuration weights generated by the optimization configuration module, including operations such as asset increase, decrease, and redistribution, to ensure that the actual operation of the digital asset management system is consistent with the optimal configuration weights. The asset execution module receives the configuration instructions from the optimization configuration module and feeds back the results to the data acquisition module after execution, forming a closed-loop system.

[0063] Refer to the attached Figure 2, which shows a flow chart of a digital asset management method based on intelligent perception provided by an embodiment of the present invention, the method comprising the following steps:

[0064] S1. Collect feature data of each dimension of digital assets, clean and standardize the collected feature data of each dimension to generate a standardized data set; calculate feature entropy based on the standardized data set; construct a feature mapping matrix based on the feature entropy; calculate the comprehensive risk score of the asset based on the feature mapping matrix;

[0065] The core information of digital assets is comprehensively collected. The collection process involves real-time monitoring of multi-dimensional feature data of digital assets, including but not limited to asset type, volatility, trading volume, holder reputation and other information. Each type of information corresponds to an independent data channel to achieve parallel collection and monitoring of different feature data, thereby improving the timeliness and accuracy of the collected and monitored feature data. For example, asset type is used to identify the basic classification attributes of assets (such as digital currency, smart contracts or virtual assets, etc.), while volatility reflects the stability and risk of the market by monitoring the price or value fluctuations of digital assets over a period of time. Trading volume presents the activity and liquidity of the asset by summarizing historical trading data, and holder reputation is a stability score comprehensively evaluated through multi-dimensional data such as historical transaction records and account behavior, which is used to characterize the holder's credit level and trading stability.

[0066] In order to ensure the consistency and reliability of feature data, the collected feature data is cleaned and standardized. Data cleaning is mainly used to remove outliers in feature data, such as extremely volatile values ​​or incomplete data records, to ensure that the feature data is accurate and representative. Subsequently, standardization is used to unify feature data from different sources into one dimension to eliminate the impact of unit differences. For example, asset volatility and trading volume are normalized to ensure that the weight distribution in subsequent algorithms is comparable, thereby forming a standardized data set.

[0067] The standardized data set is processed. Each attribute (or dimension) in the standardized data set is a feature. The contribution of each feature in the classification is obtained by calculating the feature entropy. The purpose of the feature entropy is to assign weights between different features so that decisions can be made based on the importance of each feature when classifying assets. The specific calculation formula for the feature entropy is as follows:

[0068]

[0069] Among them, feature entropy H(X) represents the importance of each feature to the classification decision, which is used to measure the uncertainty of the feature set X and the weight influence of each feature in the classification process; w iis the weight of the i-th feature, which determines the influence of each feature on the classification; x i is the value of the ith feature, representing the current measured value of the feature; n is the number of features; δ is a smoothing constant used to avoid zero values ​​when calculating logarithms to ensure operational stability; α is a weight adjustment factor used to adjust the balance between features; μ is the feature mean, which balances the distribution range of features through the root mean square method to improve the stability of classification. The purpose of calculating feature entropy is to generate a weighted feature set so that the digital asset management system can use the weight of each feature to better show the similarity and importance between features in the feature mapping process, thereby improving the accuracy of classification.

[0070] After the feature entropy calculation is completed, the feature mapping matrix is ​​constructed using the weighted results of the features. The feature mapping matrix is ​​used to describe the similarity relationship between features in a high-dimensional space, thereby providing support for the automated classification of digital assets. The construction of the feature mapping matrix is ​​based on the weight and angle difference of each pair of feature values. The specific calculation formula is:

[0071]

[0072] Among them, M ij is the feature mapping matrix, which represents the relationship between the i-th feature and the j-th feature and is used to describe the similarity between the two in multidimensional space; j is the value of the jth feature; θ ij It represents the angle between the ith feature and the jth feature, and quantifies the directional difference between the features. The feature entropy H(X) is introduced here as a weighting factor to ensure that the feature mapping matrix can fully reflect the weight of the feature in the classification decision. λ is a mapping adjustment parameter, which is used to control the nonlinear effect of the similarity between features to ensure that the appropriate similarity can be maintained when the feature value is large. The feature mapping matrix generates a high-dimensional feature similarity graph, through which digital assets can be preliminarily grouped by similarity, providing a data basis for the next step of risk analysis.

[0073] After the feature mapping matrix is ​​generated, the risk score of each type of digital asset is calculated based on the similarity information of the features to identify potential high-risk assets; a comprehensive risk score is generated by combining multiple risk factors, including market volatility, asset liquidity, and holder reputation. The specific formula for the comprehensive risk score is as follows:

[0074]

[0075] Among them, S R Represents the comprehensive risk score of an asset, which is used to quantify the risk level of various types of digital assets under market conditions; m is the number of risk factors; is the weight of the kth risk factor, calculated based on the correlation in the feature mapping matrix; M v Indicates the degree of market volatility, which is used to reflect the risk impact of the external environment on digital assets; β k is the adjustment coefficient, which is used to control the sensitivity of market fluctuations to the kth risk factor; C s Indicates the credit score of the holder, which is used to influence the stability of digital assets; L i It is the liquidity of assets, representing the circulation frequency of digital assets in the market; M ik is the similarity measure between the i-th feature and the feature associated with the k-th risk factor in the feature mapping matrix; H(x i ) is the characteristic entropy of the ith feature. The comprehensive risk scoring formula combines multiple factors to systematically quantify the risk level of different digital assets, identify potential risks, and pass the comprehensive risk score to the next step of configuration optimization as a key input.

[0076] S2. Based on the comprehensive risk score, construct an objective function and balance the return and risk by optimizing the objective function; use the adaptive weight adjustment algorithm based on Bayesian update to solve the optimal value of the optimization objective function, maximize the return of digital assets and balance risk control.

[0077] After the comprehensive risk score is determined, the configuration weight of each digital asset is dynamically adjusted based on the balance between return and risk; the maximization of the comprehensive rate of return of the asset is taken as the optimization goal, and the objective function is constructed, and the comprehensive risk score is used as a constraint in the optimization process. The specific formula of the objective function is as follows:

[0078]

[0079] Among them, J represents the objective function, and the goal is to maximize this value to balance the return and risk; is the configuration weight of the u-th type of digital assets, which is a variable dynamically adjusted during the optimization process; A u represents the current value of the u-th digital asset, which is used to measure the actual gain or loss of the digital asset; γ is the market volatility sensitivity, which is used to control the impact of market volatility on earnings; R y Represents the current rate of return, and its square root processing is used to smooth the impact of market fluctuations on returns; is a risk adjustment parameter used to balance returns and risks; N is the total number of asset categories in the current asset allocation. The optimization objective function combines the balance between returns and risks to ensure that the digital asset management system maximizes returns while keeping risks under control.

[0080] The market environment and asset fluctuations are dynamic. The risks and returns of different digital assets will change in real time with market conditions during actual operation. In order to enable the digital asset management system to adaptively optimize asset allocation under different market conditions and always maintain a balance between maximizing returns and controlling risks, it is necessary to adjust the allocation weights of digital assets based on the latest market feedback. Ensure that optimization goals remain valid and relevant to current market conditions.

[0081] In order to solve the optimal value of the optimization objective function, an adaptive weight adjustment algorithm based on Bayesian updating is adopted to update the configuration weight according to the latest observation data D (including market volatility, yield, asset liquidity, holder reputation and other information of digital assets) The distribution of configuration weights can be adapted to market changes; the optimization process is based on the gradient descent method, which gradually adjusts the configuration weights to find the optimal configuration that maximizes the objective function.

[0082] The process of solving the optimization objective function is as follows:

[0083] 1. Calculate Bayesian posterior probability Based on the latest observation data and risk score, the posterior probability of each configuration weight is calculated using the Bayesian update formula. For each asset, calculate That is, in the current weight configuration The product of the conditional probability of the latest observed data D and the prior probability is used to obtain the posterior probability of each configuration weight, which represents the configuration priority under current market conditions. The specific formula is as follows:

[0084]

[0085] in, It represents the probability of asset weight distribution under the latest observation data D, which is the optimal distribution after asset allocation optimization; Indicates that in the current weight configuration The probability of observing the latest observation data D under the current market environment is used to measure the adaptability of the configuration strategy to the current market environment; P(D) represents the overall probability of observing the latest observation data D, which is called marginal likelihood; D u is the observed data related to the u-th asset; ∈ is a smoothing constant, which is used to prevent invalid results when the data value is zero; The risk score of the u-th asset is used to dynamically adjust the risk of the allocation strategy during the Bayesian update process. Through the Bayesian update process, the weights can be dynamically updated when market fluctuations are observed, so that the optimization objective function remains in the best state of maximizing returns and controlling risks.

[0086] 2. Initialize the weight distribution: set the mean of the Bayesian posterior probability Or MAP value (maximum a posteriori probability value) is used as the initial configuration, and the weights of each configuration are initialized to the optimal configuration under the current conditions; the initial weight value is determined by the posterior distribution after Bayesian update, so that the optimization process can start from the current market status and improve the convergence efficiency.

[0087] 3. Optimize the value of the objective function J by gradient descent: After initializing the weights, based on the gradient information of the objective function J, gradually adjust each configuration weight by gradient descent To find the configuration that maximizes the objective function J. The gradient update formula is:

[0088]

[0089] in, represents the configuration weight after the iteration of time step t+1, that is, the updated configuration weight; represents the configuration weight after t time step iteration, that is, the configuration weight before update; t represents the current time step; η is the learning rate, which is used to control the step size of each update. Represents the objective function J for the configuration weight The partial derivative of is used to guide the direction of configuration weight adjustment.

[0090] 4. Iterative update and dynamic adaptation: The digital asset management system iterates the above gradient update process at each time step and recalculates the Bayesian posterior probability after each update of the configuration weight Ensure that the weight configuration always meets the current market conditions. Through multiple iterations, in the process of asymptotically approaching the optimal configuration, the configuration weight is dynamically adjusted to maximize the objective function J while strictly following the risk control conditions.

[0091] 5. Implementation of constraints: During the gradient update process, ensure that each configuration weight Satisfy non-negativity and sum constraints. Non-negativity constraint Ensure that the allocation weight of each asset is always positive to avoid negative allocation. This is enforced through existing constrained optimization methods (such as the Lagrange multiplier method) so that the total allocation is always within the available funds.

[0092] Through the above process, Bayesian updating provides the latest market feedback at each time step, allowing the digital asset management system to dynamically correct the weight distribution, thereby adapting to market fluctuations and risk conditions in real time. The optimization process combines Bayesian posterior information and gradient information in weight configuration, allowing the digital asset management system to readjust the configuration weights every time the observed data changes, forming an adaptive closed-loop optimization structure. Ultimately, through iterative updates, the balance between maximizing the benefits of digital assets and controlling risks is achieved.

[0093] In summary, a digital asset management method and system based on intelligent perception has been completed.

[0094] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0095] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0096] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such 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, and should be included in the protection scope of the present invention.

Claims

1. A digital asset management method based on intelligent perception, characterized in that: The following steps are involved: S1. Collect feature data of each dimension of digital assets, clean and standardize the collected feature data of each dimension, and generate a standardized data set; calculate feature entropy based on the standardized data set; Based on the feature entropy, construct the feature mapping matrix; Based on the feature mapping matrix, calculate the comprehensive risk score of the asset; S2. Based on the comprehensive risk score, construct an objective function and balance the return and risk by optimizing the objective function; use the adaptive weight adjustment algorithm based on Bayesian update to solve the optimal value of the optimization objective function, maximize the return of digital assets and balance risk control.

2. The digital asset management method based on intelligent perception according to claim 1 is characterized in that: The S1 specifically includes: Each dimension in the standardized data set is a feature, and the contribution of each feature in the classification is obtained by calculating the feature entropy. The specific calculation formula of the feature entropy is as follows: Among them, feature entropy H(X) represents the importance of each feature to the classification decision, which is used to measure the uncertainty of the feature set X and the weight influence of each feature in the classification process; w i is the weight of the i-th feature; x i is the value of the ith feature; n is the number of features; δ is a smoothing constant used to avoid zero values ​​when calculating logarithms; α is a weight adjustment factor; and μ is the feature mean.

3. The digital asset management method based on intelligent perception according to claim 1 is characterized in that: The S1 specifically includes: By combining the risk factors of market volatility, asset liquidity and holder reputation, a comprehensive risk score is generated. The specific formula is as follows: Among them, S R represents the comprehensive risk score of the asset; m is the number of risk factors; is the weight of the kth risk factor; M v Indicates the degree of market volatility; β k is the adjustment coefficient, which is used to control the sensitivity of market fluctuations to the kth risk factor; C s Indicates the credit score of the holder; L i It is the liquidity of assets.

4. The digital asset management method based on intelligent perception according to claim 1 is characterized in that: The S2 specifically includes: The configuration weight of each digital asset is dynamically adjusted based on the balance between returns and risks, and the maximization of the comprehensive rate of return of the asset is taken as the optimization goal. The objective function is constructed and the comprehensive risk score is used as a constraint in the optimization process. The specific formula of the objective function is as follows: Where J represents the objective function; A is the configuration weight of the u-th type of digital assets; u represents the current value of the u-th digital asset; γ is the market volatility sensitivity; R y Indicates the current rate of return; is the risk adjustment parameter; N is the total number of asset classes in the current asset allocation.

5. The digital asset management method based on intelligent perception according to claim 4 is characterized in that: The S2 specifically includes: The adaptive weight adjustment algorithm based on Bayesian updating updates the distribution of configuration weights according to the latest observation data. The optimization process of the objective function is based on the gradient descent method. By gradually adjusting the configuration weights, the optimal configuration that maximizes the objective function is found.

6. The digital asset management method based on intelligent perception according to claim 5 is characterized in that: The S2 specifically includes: The process of solving the optimization objective function includes: Step 1: Calculate the Bayesian posterior probability; Calculate the posterior probability of each configuration weight based on the latest observation data and risk score; Step 2: Initialize the weight distribution; use the mean or maximum posterior probability of the Bayesian posterior probability as the initial configuration, and initialize the weights of each configuration to the optimal configuration under the current conditions; Step 3: Optimize the value of the objective function by gradient descent. After initializing the weight distribution, gradually adjust each configuration weight by gradient descent to find the configuration that maximizes the objective function. Step 4: Iterative update and dynamic adaptation: recalculate the Bayesian posterior probability after each adjustment of the configuration weight, and dynamically adjust the configuration weight through iteration to maximize the objective function; Step 5: During the gradient update process, each configuration weight satisfies the non-negativity and sum constraints.

7. The digital asset management method based on intelligent perception according to claim 6 is characterized in that: The S2 specifically includes: Bayesian Posterior Probability The specific calculation formula is: in, Indicates that in the current weight configuration The probability of observing the latest observation data D under represents the prior probability; P(D) represents the overall probability of observing the latest observation data D; D u is the observed data related to the u-th asset; ∈ is a smoothing constant, which is used to prevent invalid results when the data value is zero; Give risk score to the u-th asset class.

8. A digital asset management system based on intelligent perception, applied to the digital asset management method based on intelligent perception as claimed in claim 1, characterized in that: Includes the following parts: Data acquisition module, data preprocessing module, feature extraction module, feature mapping matrix construction module, risk assessment module, optimization configuration module, asset execution module; The data collection module is responsible for real-time collection of various characteristic data of digital assets. Through the parallel data channel structure, various characteristic data are monitored in real time. The data collection module directly transmits the collected various characteristic data to the data preprocessing module. The data preprocessing module cleans and standardizes various feature data to generate a standardized data set; each dimension in the standardized data set is a feature; The data preprocessing module passes the standardized data set to the feature extraction module; The feature extraction module uses the standardized data set to calculate the feature entropy for each feature; The feature extraction module transfers the feature entropy of each feature to the feature mapping matrix construction module; The feature mapping matrix construction module generates a feature mapping matrix according to feature entropy; The feature mapping matrix construction module transmits the constructed feature mapping matrix to the risk assessment module; The risk assessment module calculates the comprehensive risk score of each type of asset based on the feature mapping matrix; The risk assessment module transmits the generated comprehensive risk score to the optimization configuration module; The optimization configuration module performs asset configuration optimization based on the comprehensive risk score, updates the configuration weights according to the latest market feedback through the adaptive weight adjustment algorithm based on Bayesian update, and generates configuration instructions; the optimization configuration module transmits the optimal configuration weights to the asset execution module; The asset execution module executes the asset allocation strategy according to the optimal allocation weights; The asset execution module receives the configuration instructions and feeds the results back to the data acquisition module after execution, forming a closed-loop system.

Images