Block chain optimal feature selection method and device, server and storage medium

Abstract

The invention provides a block chain optimal feature selection method and device, a server and a storage medium, and the method comprises the steps: training a reference model based on a complete data set, and evaluating the basic accuracy of the reference model; the complete data set is represented as a feature set, and multiple columns of feature data exist in the feature set; performing importance evaluation on each column of feature data in the feature set to generate a feature importance bar chart; selecting feature data with a preset proportion from the feature set as an initial feature population; and iteratively optimizing the initial feature population through a genetic algorithm, and selecting the feature population with the highest accuracy of the reference model as the optimal feature population. In the technical scheme provided by the invention, the genetic algorithm is combined with feature importance evaluation, an efficient feature screening mechanism is constructed by quantifying the contribution degree of each feature to model prediction, the complex association relationship among the features in the block chain data is effectively processed, and the calculation efficiency is improved while the model accuracy is ensured.

Description

Technical Field

[0001] This invention relates to the field of feature selection technology, and specifically to a method, apparatus, server, and storage medium for selecting optimal features in blockchain. Background Technology

[0002] With the rapid development of information technology, massive amounts of high-dimensional data are exploding in number. Correspondingly, the issues of data correlation and uncertainty have also arisen. How to effectively utilize data, reduce redundancy, ensure information integrity, and extract useful information is one of the core challenges in the field of big data mining. Feature selection, as a commonly used data preprocessing method in data mining, filters out noise and irrelevant features to select the subset of features that best reflect the essence of the data, effectively removing redundant features, reducing data dimensionality, and improving model performance.

[0003] Blockchain data is typically characterized by high dimensionality, large scale, and low value density. Directly using all features to train machine learning models leads to low computational efficiency and poor model interpretability. Traditional feature selection methods are ill-suited to the high-efficiency and accurate requirements of blockchain data and cannot effectively handle the complex relationships between features. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention provide a method, apparatus, server and storage medium for selecting optimal features in blockchain, in order to solve the problem that traditional feature selection methods in the prior art are difficult to adapt to the requirements of high efficiency and accuracy of blockchain data, and cannot effectively deal with the complex correlation between features.

[0005] In a first aspect, embodiments of the present invention provide a method for selecting optimal features in a blockchain, the method comprising:

[0006] A baseline model is trained based on a complete dataset, and the basic accuracy of the baseline model is evaluated; the complete dataset is represented as a feature set, which has multiple columns of feature data;

[0007] The importance of each column of feature data in the feature set is evaluated, and a feature importance bar chart is generated;

[0008] A predetermined proportion of feature data is selected from the feature set as the initial feature population;

[0009] The initial feature population is iteratively optimized using a genetic algorithm, and the feature population with the highest accuracy of the benchmark model is selected as the optimal feature population.

[0010] In one possible implementation, the step of evaluating the importance of each column of feature data in the feature set and generating a feature importance bar chart includes:

[0011] Each column of feature data in the feature set is shuffled sequentially. The perturbation accuracy of the benchmark model is evaluated based on the perturbed feature set to calculate the importance of that column of feature data. This process continues until the importance of each column of feature data in the feature set has been evaluated. Finally, a feature importance bar chart is generated based on the importance of each column of feature data.

[0012] In one possible implementation, the importance of the feature data is the difference between the base accuracy and the perturbation accuracy.

[0013] In one possible implementation, the lower limit of the preset ratio is 0.4 and the upper limit is 0.6.

[0014] In one possible implementation, the step of iteratively optimizing the initial feature population using a genetic algorithm to select the feature population with the highest accuracy of the benchmark model as the optimal feature population includes:

[0015] Based on the feature importance bar chart, a selection operation is performed on the initial feature population to select a portion of the feature data and generate a selected population.

[0016] Perform a splitting operation on the selected population to obtain a split population;

[0017] Perform a probabilistic mutation operation on the split population to obtain the mutated population;

[0018] The accuracy of the baseline model is calculated based on the mutated population, and iterative optimization is performed until the maximum number of iterations is reached. The feature population with the highest accuracy is selected from all iteration rounds as the optimal feature population.

[0019] In one possible implementation, before calculating the accuracy of the baseline model based on the mutated population and iteratively optimizing it until the maximum number of iterations is reached, and selecting the feature population with the highest accuracy from all iteration rounds as the optimal feature population, the method further includes:

[0020] Determine whether the number of features in the mutated population is less than a first preset threshold;

[0021] If it is determined that the number of features is less than the first preset threshold, then the number of features is increased to the second preset threshold; the second preset threshold is greater than the first preset threshold.

[0022] If it is determined that the number of features is greater than the first preset threshold, then the step of calculating the accuracy of the benchmark model based on the mutated population and iteratively optimizing it is performed until the maximum number of iterations is reached, and the feature population with the highest accuracy is selected as the optimal feature population from all iteration rounds is selected.

[0023] In one possible implementation, the probability of selecting the feature data during the selection operation is proportional to the importance of the feature data.

[0024] Secondly, embodiments of the present invention provide a blockchain optimal feature selection device, the device comprising:

[0025] The training module is used to train a benchmark model based on a complete dataset and evaluate the accuracy of the benchmark model; the complete dataset is represented as a feature set, which has multiple columns of feature data.

[0026] The evaluation module is used to evaluate the importance of each column of feature data in the feature set and generate a feature importance bar chart;

[0027] The selection module is used to select a preset proportion of feature data from the feature set as an initial feature population;

[0028] The iterative module is used to iteratively optimize the initial feature population using a genetic algorithm, and select the feature population with the highest accuracy of the benchmark model as the optimal feature population.

[0029] Thirdly, embodiments of the present invention provide a server, the server including one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs include instructions that, when executed by the server, cause the server to perform the blockchain optimal feature selection method as described in the first aspect or any possible implementation of the first aspect.

[0030] Fourthly, embodiments of the present invention provide a computer-readable storage medium comprising a stored program, wherein, when the program is running, it controls the device where the computer-readable storage medium is located to execute the blockchain optimal feature selection method as described in the first aspect or any possible implementation thereof.

[0031] The technical solution provided in this invention addresses the characteristics of blockchain data, such as high dimensionality, large scale, and low value density. It combines genetic algorithms with feature importance assessment, quantifies the contribution of each feature to the model prediction, and constructs an efficient feature selection mechanism. This mechanism can effectively handle the complex relationships between features in blockchain data, significantly improve computational efficiency while ensuring model accuracy, and overcome the problem of insufficient adaptability of traditional methods in blockchain scenarios.

[0032] In this embodiment of the invention, feature data in the feature set is systematically shuffled to observe the degree of model performance degradation, thus accurately quantifying feature importance. Simultaneously, selection, splitting, and mutation operations using a genetic algorithm enable intelligent exploration of the feature space, improving the search efficiency for the optimal feature population. Feature importance, as a core indicator of fitness evaluation, guides the evolutionary direction.

[0033] In this embodiment of the invention, a complete process of benchmark model training and evaluation, feature importance assessment, and genetic algorithm feature selection is constructed. Each step in the above process generates traceable evaluation records, which is applicable to blockchain scenarios such as cryptocurrency transaction behavior analysis and cross-chain transaction monitoring. It provides a transparent and verifiable feature selection solution for blockchain data feature mining, and has high practical value and promotion prospects. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating a method for selecting optimal blockchain features according to an embodiment of the present invention.

[0035] Figure 2 A flowchart illustrating another method for selecting optimal blockchain features provided in an embodiment of the present invention.

[0036] Figure 3 This is a schematic diagram of a blockchain optimal feature selection device provided in an embodiment of the present invention.

[0037] Figure 4 This is a schematic diagram of a server provided in an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of this invention clearer and more understandable, the invention will be further described below in conjunction with the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0039] Figure 1This is a flowchart illustrating a method for selecting optimal features in a blockchain according to an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0040] Step 101: Train a baseline model based on the complete dataset and evaluate the baseline model's basic accuracy; the complete dataset is represented as a feature set, which contains multiple columns of feature data.

[0041] Step 102: Evaluate the importance of each column of feature data in the feature set and generate a feature importance bar chart.

[0042] Step 103: Select a preset proportion of feature data from the feature set as the initial feature population.

[0043] Step 104: Iteratively optimize the initial feature population using a genetic algorithm, and select the feature population with the highest accuracy of the benchmark model as the optimal feature population.

[0044] The technical solution provided in this invention addresses the characteristics of blockchain data, such as high dimensionality, large scale, and low value density. It combines genetic algorithms with feature importance assessment, quantifies the contribution of each feature to the model prediction, and constructs an efficient feature selection mechanism. This mechanism can effectively handle the complex relationships between features in blockchain data, significantly improve computational efficiency while ensuring model accuracy, and overcome the problem of insufficient adaptability of traditional methods in blockchain scenarios.

[0045] Figure 2 A flowchart illustrating another method for selecting optimal blockchain features provided in an embodiment of the present invention is shown below. Figure 2 As shown, the method includes:

[0046] Step 201: Train a baseline model based on the complete dataset and evaluate the baseline model's basic accuracy; the complete dataset is represented as a feature set, which contains multiple columns of feature data.

[0047] In this step, a complete dataset is formed based on blockchain transaction data, and the feature set is represented as F = {f1, f2, ..., f...} n The feature set has n columns of feature data, and each column contains multiple feature values. Model accuracy is the proportion of samples correctly classified by the baseline model out of the total number of samples. A correctly classified sample is one whose predicted class matches the true class (i.e., TP and TN).

[0048] Optionally, by combining the confusion matrix, examine the distribution of TP (actually positive, also predicted as positive), TN (actually negative, also predicted as negative), FP (actually negative, but predicted as positive), and FN (actually positive, but predicted as negative) to confirm whether the positive and negative samples are balanced, ensuring that the high accuracy is due to model performance rather than sample class imbalance.

[0049] Step 202: Sequentially shuffle each column of feature data in the feature set, evaluate the perturbation accuracy of the benchmark model based on the perturbed feature set, and calculate the importance of the feature data in that column; until the importance of each column of feature data in the feature set has been evaluated, generate a feature importance bar chart based on the importance of each column of feature data.

[0050] In this step, shuffling a column of feature data in the feature set means randomly shuffling multiple feature values ​​in that column to ensure that changes in model performance are only related to that column of feature data, accurately quantifying its importance. The importance of a feature is the difference between the base accuracy and the perturbation accuracy, i.e., the performance degradation of the baseline model. The feature importance bar chart quantifies the contribution of each abstract column of feature data to the model's prediction, describing the importance of each column, avoiding subjective bias in feature selection, and providing an intuitive and visual description of feature importance.

[0051] It should be noted that after assessing the importance of one column of feature data, the feature data in that column is restored to its original state before assessing the importance of the next column of feature data. In other words, when assessing the importance of a column of feature data, only the feature data in that column is shuffled, while the feature data in the other columns remain unchanged.

[0052] For example, the feature set includes three columns of feature data: transaction amount, time interval between adjacent transactions, and number of transactions. Each column contains multiple feature values. First, the feature values ​​in the column containing transaction amount are randomly shuffled, the importance of the transaction amount feature is assessed, and then the feature data in that column is restored. Next, the feature values ​​in the column containing time interval between adjacent transactions are randomly shuffled, the importance of the time interval feature is assessed, and then the feature data in that column is restored. Finally, the feature values ​​in the column containing number of transactions are randomly shuffled, and the importance of the number of transactions feature is assessed. At this point, the importance of each column of feature data in the feature set has been assessed, and a feature importance bar chart is generated based on the importance of each column.

[0053] Step 203: Select a preset proportion of feature data from the feature set as the initial feature population.

[0054] In this step, the preset ratio has a lower limit of 0.4 and an upper limit of 0.6. That is, the initial characteristic population.

[0055] Step 204: Based on the feature importance bar chart, perform a selection operation on the initial feature population, select some feature data to generate the selected population.

[0056] In this step, during the selection operation, the probability of selecting feature data is directly proportional to the importance of the feature data. That is, the higher the importance of the feature data, the greater the probability of it being selected into the selected population; the lower the importance of the feature data, the smaller the probability of it being selected into the selected population.

[0057] Step 205: Perform a splitting operation on the selected population to obtain the split population.

[0058] In this step, the splitting operation refers to splitting a single feature into multiple features to increase feature diversity.

[0059] Step 206: Perform a probabilistic mutation operation on the split population to obtain the mutated population.

[0060] In this step, mutation refers to changing one trait into another to increase trait diversity. Probabilistic mutation means that a trait in the resulting population has a chance of mutating into another trait, rather than all traits in the population mutating into another trait.

[0061] Step 207: Determine whether the number of features in the mutated population is less than the first preset threshold; if the number of features is less than the first preset threshold, proceed to step 208; if the number of features is greater than the first preset threshold, proceed to step 209.

[0062] Step 208: Supplement the number of features in the mutated population to the second preset threshold, where the second preset threshold is greater than the first preset threshold.

[0063] For example, the first preset threshold is 0.4n, which is 40% of the feature data in the feature set, and the second preset threshold is 0.6n, which is 60% of the feature data in the feature set. By introducing an adaptive adjustment mechanism for the number of features, when the number of features is less than the first preset threshold, the number of features is automatically supplemented to the second preset threshold. This maintains the diversity of the evolutionary process while ensuring the search efficiency of the optimal feature population.

[0064] Step 209: Calculate the accuracy of the baseline model based on the mutated population and iteratively optimize it until the maximum number of iterations is reached. Select the feature population with the highest accuracy from all iteration rounds as the optimal feature population.

[0065] In this step, the parameter t represents the iteration round, and the maximum number of iterations is T. That is, the range of the iteration rounds is 1 to T. If the maximum number of iterations has not been reached, then proceed to step 204.

[0066] In related technologies, blockchains typically store a large number of redundant features. For example, different fields may represent the same core information, differing only in data format or storage location; or different data tables may store corresponding block data, differing only in their purpose. In this embodiment of the invention, by quantifying the contribution of each feature to the model's prediction and selecting features based on that contribution, redundant features can be efficiently and accurately removed, data dimensionality reduced, and model performance improved.

[0067] The technical solution provided in this invention addresses the characteristics of blockchain data, such as high dimensionality, large scale, and low value density. It combines genetic algorithms with feature importance assessment, quantifies the contribution of each feature to the model prediction, and constructs an efficient feature selection mechanism. This mechanism can effectively handle the complex relationships between features in blockchain data, significantly improve computational efficiency while ensuring model accuracy, and overcome the problem of insufficient adaptability of traditional methods in blockchain scenarios.

[0068] In this embodiment of the invention, feature data in the feature set is systematically shuffled to observe the degree of model performance degradation, thus accurately quantifying feature importance. Simultaneously, selection, splitting, and mutation operations using a genetic algorithm enable intelligent exploration of the feature space, improving the search efficiency for the optimal feature population. Feature importance, as a core indicator of fitness evaluation, guides the evolutionary direction.

[0069] In this embodiment of the invention, a complete process of benchmark model training and evaluation, feature importance assessment, and genetic algorithm feature selection is constructed. Each step in the above process generates traceable evaluation records, which is applicable to blockchain scenarios such as cryptocurrency transaction behavior analysis and cross-chain transaction monitoring. It provides a transparent and verifiable feature selection solution for blockchain data feature mining, and has high practical value and promotion prospects.

[0070] Figure 3 This is a schematic diagram of a blockchain optimal feature selection device provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the device includes a training module 11, an evaluation module 12, a selection module 13, and an iteration module 14. The training module trains a baseline model based on a complete dataset and evaluates the accuracy of the baseline model; the complete dataset is represented as a feature set, which contains multiple columns of feature data. The evaluation module evaluates the importance of each column of feature data in the feature set and generates a feature importance bar chart. The selection module selects a predetermined proportion of feature data from the feature set as the initial feature population. The iteration module iteratively optimizes the initial feature population using a genetic algorithm, selecting the feature population with the highest accuracy of the baseline model as the optimal feature population.

[0071] In this embodiment of the invention, the evaluation module 12 is specifically used to sequentially shuffle each column of feature data in the feature set, evaluate the perturbation accuracy of the benchmark model based on the perturbated feature set, and calculate the importance of the column of feature data; until the importance of each column of feature data in the feature set has been evaluated, a feature importance bar chart is generated based on the importance of each column of feature data.

[0072] In this embodiment of the invention, the iteration module 14 includes a selection submodule, a splitting submodule, a mutation submodule, and an iteration submodule. The selection submodule performs a selection operation on the initial feature population based on a feature importance bar chart, selecting a subset of feature data to generate a selected population. The splitting submodule performs a splitting operation on the selected population to obtain a split population. The mutation submodule performs a probabilistic mutation operation on the split population to obtain a mutated population. The iteration submodule calculates the accuracy of the baseline model based on the mutated population and iteratively optimizes it until the maximum number of iterations is reached, selecting the feature population with the highest accuracy from all iteration rounds as the optimal feature population.

[0073] In this embodiment of the invention, the iteration module 14 further includes a judgment submodule and a supplementary submodule. The judgment submodule is used to determine whether the number of features in the mutated population is less than a first preset threshold; if the judgment submodule determines that the number of features is less than the first preset threshold, then the supplementary submodule is triggered to supplement the number of features to a second preset threshold; the second preset threshold is greater than the first preset threshold; if the judgment submodule determines that the number of features is greater than the first preset threshold, then the iteration submodule is triggered to perform the step of calculating the accuracy of the benchmark model based on the mutated population, and iteratively optimizing until the maximum number of iterations is reached, and selecting the feature population with the highest accuracy from all iteration rounds as the optimal feature population.

[0074] The technical solution provided in this invention addresses the characteristics of blockchain data, such as high dimensionality, large scale, and low value density. It combines genetic algorithms with feature importance assessment, quantifies the contribution of each feature to the model prediction, and constructs an efficient feature selection mechanism. This mechanism can effectively handle the complex relationships between features in blockchain data, significantly improve computational efficiency while ensuring model accuracy, and overcome the problem of insufficient adaptability of traditional methods in blockchain scenarios.

[0075] In this embodiment of the invention, feature data in the feature set is systematically shuffled to observe the degree of model performance degradation, thus accurately quantifying feature importance. Simultaneously, selection, splitting, and mutation operations using a genetic algorithm enable intelligent exploration of the feature space, improving the search efficiency for the optimal feature population. Feature importance, as a core indicator of fitness evaluation, guides the evolutionary direction.

[0076] In this embodiment of the invention, a complete process of benchmark model training and evaluation, feature importance assessment, and genetic algorithm feature selection is constructed. Each step in the above process generates traceable evaluation records, which is applicable to blockchain scenarios such as cryptocurrency transaction behavior analysis and cross-chain transaction monitoring. It provides a transparent and verifiable feature selection solution for blockchain data feature mining, and has high practical value and promotion prospects.

[0077] This invention provides a computer-readable storage medium, which includes a stored program. When the program runs, it controls the device where the computer-readable storage medium is located to execute the steps of the above-described embodiment of the optimal blockchain feature selection method. For a detailed description, please refer to the above-described embodiment of the optimal blockchain feature selection method.

[0078] Figure 4 A schematic diagram of a server provided in an embodiment of the present invention, as shown below. Figure 4 As shown, the server 3 in this embodiment includes a processor 301, a memory 302, and a computer program 303 stored in the memory 302 and executable on the processor 301. When the processor 301 executes the computer program 303, it implements the steps in the various method embodiments described above. Alternatively, when the processor 301 executes the computer program 303, it implements the functions of each module / unit in the various device embodiments described above.

[0079] Server 3 can be an electronic device such as a desktop computer, laptop, handheld computer, or cloud server. Server 3 may include, but is not limited to, processor 301 and memory 302. Those skilled in the art will understand that... Figure 4 This is merely an example of server 3 and does not constitute a limitation on server 3. It may include more or fewer components than shown in the illustration, or different components.

[0080] The processor 301 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0081] The memory 302 can be an internal storage unit of the server 3, such as a hard drive or memory of the server 3. The memory 302 can also be an external storage device of the server 3, such as a plug-in hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, or FlashCard. The memory 302 can also include both internal and external storage units of the server 3. The memory 302 is used to store computer programs and other programs and data required by electronic devices.

[0082] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0083] The above description is merely a preferred 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 blockchain optimal feature selection method, characterized in that, The method comprises: training a benchmark model based on a complete data set, evaluating the basic accuracy of the benchmark model; the complete data set is represented as a feature set, and the feature set has multiple column feature data; performing importance evaluation on each column feature data in the feature set to generate a feature importance bar chart; selecting a preset proportion of feature data from the feature set as an initial feature population; iteratively optimizing the initial feature population by a genetic algorithm, and selecting the feature population with the highest accuracy of the benchmark model as an optimal feature population.

2. The method of claim 1, wherein, The importance evaluation on each column feature data in the feature set to generate a feature importance bar chart comprises: sequentially disturbing each column feature data in the feature set, evaluating the disturbance accuracy of the benchmark model based on the disturbed feature set to calculate the importance of the column feature data; until each column feature data in the feature set is evaluated for importance, a feature importance bar chart is generated based on the importance of each column feature data.

3. The method of claim 2, wherein, The importance of the feature data is the difference between the basic accuracy and the disturbance accuracy.

4. The method of claim 1, wherein, The lower limit of the preset proportion is 0.4, and the upper limit is 0.

6.

5. The method of claim 1, wherein, The iteratively optimizing the initial feature population by a genetic algorithm, and selecting the feature population with the highest accuracy of the benchmark model as an optimal feature population comprises: performing a selection operation on the initial feature population based on the feature importance bar chart, selecting part of the feature data to generate a selected population; performing a splitting operation on the selected population to obtain a split population; performing a probabilistic mutation operation on the split population to obtain a mutated population; calculating the accuracy of the benchmark model based on the mutated population, and iteratively optimizing until a maximum iteration number is reached, and selecting the feature population with the highest accuracy from all iteration rounds as an optimal feature population.

6. The method of claim 5, wherein, Before the calculating the accuracy of the benchmark model based on the mutated population, and iteratively optimizing until a maximum iteration number is reached, and selecting the feature population with the highest accuracy from all iteration rounds as an optimal feature population, the method further comprises: determining whether the number of features in the mutated population is less than a first preset threshold; if it is determined that the number of features is less than the first preset threshold, the number of features is supplemented to a second preset threshold; the second preset threshold is greater than the first preset threshold; if it is determined that the number of features is greater than the first preset threshold, the step of calculating the accuracy of the benchmark model based on the mutated population, and iteratively optimizing until a maximum iteration number is reached, and selecting the feature population with the highest accuracy from all iteration rounds as an optimal feature population is performed.

7. The method of claim 5, wherein, In the selection operation, the probability of selecting the feature data is proportional to the importance of the feature data. 8.A blockchain optimal feature selection device, characterized in that, The device comprises: a training module configured to train a benchmark model based on a complete data set, and evaluate the accuracy of the benchmark model; the complete data set is represented as a feature set, and the feature set has multiple column feature data; An evaluation module is configured to evaluate the importance of each column of feature data in the feature set and generate a feature importance bar chart; A selection module is configured to select a preset proportion of feature data from the feature set as an initial feature population; An iteration module is configured to perform iterative optimization on the initial feature population by using a genetic algorithm, and select a feature population with the highest accuracy of the benchmark model as an optimal feature population.

9. A server, characterized by The server comprises one or more processors; a memory; and one or more computer programs, wherein the one or more computer programs are stored in the memory, and the one or more computer programs comprise instructions which, when executed by the server, cause the server to perform the blockchain optimal feature selection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a stored program, wherein the program controls the device where the computer readable storage medium is located to perform the blockchain optimal feature selection method according to any one of claims 1 to 7 when the program is running.

Images