Cloud computing business informatization system platform based on block chain

Abstract

The invention discloses a cloud computing business informatization system platform based on a block chain, and the platform comprises a data collection and preprocessing module which is used for collecting and preprocessing business data, and constructing a business feature matrix; the feature compression modeling module is used for extracting spatial locality and sequence features of the commercial feature matrix based on a multilayer convolution structure; the loss function optimization module is used for constructing a loss function based on the initial potential feature expression and the reconstructed sample; the behavior modeling and performance evaluation module is used for calculating a user behavior prediction error based on the potential feature set, and constructing a multi-target performance evaluation result driven by a block chain in combination with a contract execution delay index and a service quality index in the platform; the intelligent optimization module is used for executing a seagull optimization algorithm, initializing seagull individual weights, constructing a multi-target fitness function, simulating three-stage trajectory behaviors and carrying out collaborative optimization on a model structure and scheduling strategy parameters; the model deployment module is used for deploying the optimized feature compression model to a cloud computing platform based on a block chain; and the self-adaptive updating module is used for updating the commercial feature compression model in real time based on the operation state data. According to the invention, accurate cloud computing business informatization is provided for the user.

Description

Technical Field

[0001] The present invention relates to the field of business intelligence optimization technology, and in particular to a cloud computing business information system platform based on blockchain. Background Art

[0002] Currently, with the continuous development of cloud computing, big data, and artificial intelligence technologies, the trend of digital transformation in enterprises is becoming increasingly prominent. As the core platform for resource scheduling and operational decision-making, business information systems are gradually evolving from traditional information management systems to cloud service architectures capable of real-time data processing, intelligent analysis, and automated execution. Especially in a highly concurrent and dynamically changing business environment, improving customer behavior awareness, increasing resource allocation efficiency, and enhancing the intelligence of system responses have become key issues that all business platforms urgently need to address.

[0003] In recent years, blockchain technology, owing to its decentralized, traceable, and tamper-proof properties, has been widely adopted in business scenarios such as financial transactions, supply chain management, and smart contract execution. Blockchain-based business systems offer highly reliable data sharing capabilities, providing a data foundation for cross-entity business collaboration. However, in the process of integrating with cloud computing, existing blockchain-based business information platforms often suffer from inefficient computing resource scheduling, weak behavioral modeling capabilities, and low levels of business automation. Traditional business systems primarily rely on static, rule-driven data processing processes, lacking the ability to perceive dynamic changes in user behavior and unable to intelligently optimize complex business logic.

[0004] In terms of data modeling, traditional methods often use linear dimensionality reduction, feature selection, or static clustering techniques, such as principal component analysis and K-means clustering. These methods have difficulty processing the large number of nonlinear features and high-dimensional redundant structures present in on-chain business data, limiting their adaptability in actual high-complexity business scenarios. At the same time, most existing systems do not introduce deep representation structures for on-chain behavioral data and are unable to effectively extract the behavioral patterns and resource scheduling rules hidden behind complex data. In addition, although some business intelligence systems have attempted to use deep neural networks for data modeling, due to the lack of effective latent space constraints and interpretability, their training results often have strong "black box" characteristics, which is not conducive to the rule embedding and trusted execution of on-chain business at the smart contract layer.

[0005] In terms of resource optimization, existing commercial cloud systems typically use greedy strategies, static threshold scheduling, or single-objective heuristic algorithms for resource allocation, lacking comprehensive awareness of multi-dimensional business indicators. When faced with complex scenarios such as on-chain business emergencies, contract execution delays, or abnormal user behavior, scheduling strategies are prone to falling into local optimality, making it difficult to achieve a dynamic balance between resource utilization, task execution efficiency, and service response quality. Furthermore, traditional optimization algorithms fail to fully consider the coupling between historical behavior prediction results and the current model structure, lacking a structural feedback mechanism based on prediction consistency. This results in a delayed response of scheduling optimization strategies to system state changes and a lack of generalization and continuous evolution capabilities.

[0006] In terms of system deployment, many current platforms still separate model training from model execution, making it difficult for optimized parameters from the training process to truly take effect at runtime. Furthermore, model deployment is often not deeply integrated with blockchain smart contract systems, making it difficult for model outputs to directly participate in on-chain rule-making and resource control. This results in a low level of intelligent business processes and a lack of end-to-end automated processing capabilities. Regarding on-chain operational execution, current systems often rely on preset rules for resource scheduling and service selection, lacking dynamic contract triggering mechanisms based on behavioral prediction results. This limits the platform's adaptability and execution efficiency in a volatile business environment.

[0007] In terms of model iteration, existing commercial information systems often rely on periodic offline retraining, failing to dynamically update model structure parameters based on real-time feedback. This results in insufficient timeliness and adaptability in practical applications. Especially in the context of rapidly growing on-chain business data volumes and frequently changing behavioral patterns, the lack of an adaptive closed-loop mechanism driven by operational status data will further weaken the system's intelligent response capabilities.

[0008] Therefore, how to provide a cloud computing business information system platform based on blockchain is an urgent problem that technical personnel in this field need to solve. Summary of the Invention

[0009] One purpose of the present invention is to propose a cloud computing business information system platform based on blockchain. The present invention combines a convolutional variational autoencoder with the Seagull optimization algorithm to achieve deep compression modeling of multi-source business data, intelligent reasoning of on-chain behaviors, and adaptive optimization scheduling of cloud resources on the basis of ensuring that data cannot be tampered with and business is trusted. By constructing a feature compression model driven by latent space expression and combining it with an intelligent optimization mechanism guided by multi-objective performance evaluation results, the system's data processing capabilities, behavior prediction accuracy, and resource scheduling efficiency in a high-dimensional on-chain business environment are improved.

[0010] A cloud computing business information system platform based on blockchain according to an embodiment of the present invention includes:

[0011] Data collection and preprocessing module, used to collect and preprocess business data and build a business feature matrix;

[0012] Feature compression modeling module, used to extract spatial locality and sequence features of the business feature matrix based on a multi-layer convolutional structure;

[0013] The loss function optimization module is used to construct the loss function based on the initial latent feature expression and the reconstructed sample;

[0014] The behavior modeling and performance evaluation module is used to calculate the user behavior prediction error based on the potential feature set, and to construct a blockchain-driven multi-objective performance evaluation result by combining the contract execution delay indicator and service quality indicator in the platform;

[0015] Intelligent optimization module, used to execute the seagull optimization algorithm, initialize the individual seagull weights, construct a multi-objective fitness function, simulate the three-stage trajectory behavior, and coordinately optimize the model structure and scheduling strategy parameters;

[0016] The model deployment module is used to deploy the optimized feature compression model to the blockchain-based cloud computing platform;

[0017] The adaptive update module is used to update the commercial feature compression model in real time based on the operating status data.

[0018] Optionally, modules can be connected using the following methods:

[0019] S1. Collect and pre-process business data to construct a business feature matrix;

[0020] S2. Convolutional variational autoencoders are used to build a business feature compression model. The multi-layer convolutional structure extracts the spatial locality and sequence characteristics of the business feature matrix, maps high-dimensional on-chain business data to the latent space, and generates initial latent feature expressions and reconstruction samples.

[0021] S3. Construct a loss function based on the initial latent feature expression and the reconstructed samples, and output a latent feature set for business behavior modeling and resource scheduling.

[0022] S4. Calculate the user behavior prediction error based on the potential feature set, and construct a blockchain-driven multi-objective performance evaluation result by combining the contract execution delay indicator and the service quality indicator;

[0023] S5. Using the Seagull optimization algorithm, we optimized the commercial feature compression model based on the multi-objective performance evaluation results, simulated the three-stage trajectory pattern of the Seagull attack maneuvering behavior, and performed collaborative global search optimization.

[0024] S6. Deploy the optimized business feature compression model in the cloud computing platform operating environment supported by blockchain, drive business automation processing based on behavior prediction results, and use on-chain logic to control behavior reasoning and call resource execution strategies;

[0025] S7. Collect operating status data during operation, and update the business feature compression model in real time based on the operating status data.

[0026] Optionally, the commercial business data includes order transaction information, customer behavior logs, smart contract execution records, service access records, marketing activity data, cloud resource usage, and on-chain ledger interaction data.

[0027] Optionally, the operating status data includes business response logs, cloud resource scheduling feedback information and prediction deviation indicators.

[0028] Optionally, S2 includes the following specific steps:

[0029] S21. Perform multi-scale convolution processing on the business feature matrix, using multiple convolution kernels to extract the local correlation features of the on-chain business data in the spatial structure and generate a spatial feature representation;

[0030] S22, performing batch normalization and nonlinear activation transformation on the spatial feature representation in sequence to obtain normalized spatial features;

[0031] S23. Perform sliding window modeling on the normalized spatial features in chronological order, extract the temporal change pattern of the on-chain business data, and generate a temporal feature representation;

[0032] S24, inputting the temporal feature representation into the encoder unit of the variational autoencoder, mapping it to the latent space through the probability distribution reparameterization operation, and generating an initial latent feature expression;

[0033] S25, inputting the initial latent feature expression into the decoder unit of the variational autoencoder to generate a reconstructed feature representation;

[0034] S26. Perform structural alignment and shape matching on the reconstructed feature representation and the commercial feature matrix to generate a reconstructed sample.

[0035] Optionally, S3 includes the following specific steps:

[0036] S31, aligning the shapes of the initial latent feature expression and the reconstructed samples with the commercial feature matrix to construct a reconstruction comparison tensor set;

[0037] S32, calculating a reconstruction error tensor based on the reconstruction comparison tensor set by element-wise difference;

[0038] S33, modeling the mean and variance of the distribution of the initial latent feature expression in the latent space, and constructing a distribution deviation tensor from the standard multidimensional normal distribution;

[0039] S34. Jointly model the reconstruction error tensor and the distribution deviation tensor to construct a joint loss function:

[0040]

[0041] Among them, L total represents the joint loss function value, Represents the original eigenvalue of the mth dimension of the nth sample in the input commercial feature matrix, Represents the feature value of the mth dimension of the nth sample reconstructed by the decoder, N represents the total number of input samples, and M represents the number of feature dimensions in each sample. represents the square of the mean of the nth sample on the kth latent dimension, represents the square of the standard deviation of the nth sample on the kth potential dimension, K represents the total number of dimensions of the latent space, n represents the input sample index, m represents the input feature dimension index, and k represents the latent space dimension index;

[0042] S35. Apply the business feature compression model to the business feature matrix input, use the encoder to obtain the final latent feature expression, and generate a latent feature set for business behavior modeling and resource scheduling.

[0043] Optionally, S4 includes the following specific steps:

[0044] S41. Pass the potential feature set as input to the on-chain user behavior inference structure, which includes a feature encoding unit and a time-dependency modeling unit to construct an on-chain behavior trend modeling path;

[0045] S42. Pair the potential feature set with the historical user behavior records on the platform in a time series manner, input the data into the on-chain user behavior inference structure, and output the user behavior prediction result tensor through sequence pattern modeling and evolution trend generation mechanism;

[0046] S43. Compare the user behavior prediction result tensor with the actual on-chain behavior record and calculate the prediction error tensor:

[0047]

[0048] Among them, M predict Prediction consistency metric, N represents the total number of input samples, n represents the input sample index, represents the joint loss function value of the nth sample, Represents the prediction error of the nth sample;

[0049] S44. Extract the smart contract execution records corresponding to the prediction samples from the blockchain platform, calculate the task confirmation delay and transaction settlement delay indicators in the contract process based on the prediction consistency measurement value, and generate a contract execution delay indicator set;

[0050] S45. Combine the platform service operation log to obtain the service call success rate, response time and resource utilization efficiency corresponding to the prediction sample, and generate a set of service quality indicators;

[0051] S46. Normalize and vector-fuse the contract execution delay indicator set and the service quality indicator set to generate a multi-objective performance evaluation result.

[0052] Optionally, S5 includes the following specific steps:

[0053] S51. Based on the Seagull Algorithm, the Seagull Optimization Algorithm introduces a dynamic initialization mechanism based on behavior prediction consistency measurement and potential distribution structure. It differentially weights the initial population through a prediction stability weighting strategy, constructs the importance weight vector of the individual Seagulls, initializes the Seagull individuals of the Seagull Optimization Algorithm, and sets the initial solution of each Seagull individual as a joint encoding of a set of commercial feature compression model parameters and cloud resource scheduling strategy parameters:

[0054]

[0055] Among them, W (i) represents the initial importance weight of the i-th seagull individual, represents the behavior prediction consistency measure of the nth sample, i represents the seagull individual index, μ n,k represents the mean encoder output of the nth sample on the kth potential dimension, n represents the input sample index, σ n,k represents the standard deviation of the nth sample on the kth potential dimension, μ n,k represents the mean of the nth sample on the kth latent dimension, K represents the total number of dimensions of the latent space, and k represents the dimension index of the latent space;

[0056] S52. Perform fitness evaluation on each individual seagull. After normalizing the three indicators of prediction error, contract delay, and service quality corresponding to each individual seagull, they are weighted and summed using weight coefficients of 0.5, 0.3, and 0.2, respectively, to generate a comprehensive fitness score.

[0057] S53. Based on the position guidance mechanism and nonlinear contraction update mechanism for the seagull with the highest comprehensive fitness score, the trajectories of all seagulls are adjusted to simulate the three-stage pattern of seagull attack and maneuver behavior, including the initial convergent flight stage, the mid-term circling stage, and the final precise attack stage, to generate an updated solution set.

[0058] S54, based on the updated solution set, re-evaluate the fitness value of the seagull individuals and update them according to the global optimal retention mechanism until the maximum number of iterations 100 is met to generate the optimal solution group;

[0059] S55. Decode the optimal solution into commercial feature compression model structure parameters, and output the optimized commercial feature compression model configuration.

[0060] Optionally, S6 includes the following specific steps:

[0061] S61. Deploy the optimized business feature compression model to the feature processing unit in the blockchain-based cloud computing platform to perform real-time access and compression processing of the on-chain business data;

[0062] S62. Based on the potential feature tensor output by the compression model, behavior prediction is performed to generate a set of prediction results for user behavior, resource consumption trends, and contract fulfillment likelihood;

[0063] S63. Based on the set of prediction results, match the preset on-chain policy template and generate a business automation process configuration;

[0064] S64. Write the business automation process configuration into the smart contract trigger mechanism in the blockchain platform, and control the behavior reasoning call path corresponding to the behavior prediction result through the on-chain logic;

[0065] S65. Based on the behavioral reasoning call path, the corresponding resource execution policy configuration is read from the chain, and the relevant computing nodes in the cloud platform are triggered to complete resource allocation, task scheduling and processing response.

[0066] Optionally, S7 includes the following specific steps:

[0067] S71. During the operation of the blockchain-based cloud computing platform, collect operational status data during business execution;

[0068] S72. Perform data cleaning and structural processing on the operating status data to generate a standardized status data set;

[0069] S73. Perform feature alignment processing on the normalized state dataset and the initial business feature matrix to generate an incremental training dataset containing historical features and current feedback;

[0070] S74. Input the incremental training data set into the deployed commercial feature compression model, execute the micro-batch incremental training process, and generate an intermediate updated commercial feature compression model parameter set;

[0071] S75. Perform stability evaluation and performance verification on the intermediate updated commercial feature compression model parameter set to determine whether the preset update conditions are met. If so, replace the current commercial feature compression model parameters.

[0072] S76. Redeploy the replaced business feature compression model to the cloud computing platform operating environment, continue to receive on-chain business data and behavior prediction inputs, and form an adaptive closed-loop update mechanism for the business feature compression model.

[0073] The beneficial effects of the present invention are:

[0074] In terms of business data modeling, this paper introduces a convolutional variational autoencoder as a feature compression model, combining multi-scale convolution with a time series modeling mechanism to effectively extract spatial local features and behavioral temporal features from high-dimensional on-chain business data, significantly improving the compactness and discriminability of feature expression. Through the reparameterized sampling process in the latent space, this paper not only retains the core patterns in the input data, but also constructs a low-dimensional latent feature set that can be used for on-chain contract reasoning and predictive calculations, providing a structurally stable and highly expressive input foundation for subsequent behavioral modeling and resource scheduling.

[0075] In terms of multi-objective optimization, this invention integrates behavior prediction error, contract execution delay, and service quality indicators to construct a multidimensional performance vector. This vector guides the joint evolution of the system model based on the Seagull optimization algorithm. By introducing a behavior prediction consistency metric and a latent space structure perception mechanism, individual initialization weights, and a dynamic trajectory adjustment strategy, this effectively enhances the optimization algorithm's global search capability and local convergence speed, overcoming the vulnerability of traditional single-objective optimization to local optimality. Leveraging a three-stage trajectory model, this invention dynamically adjusts the compression model structure and scheduling strategy parameters based on real-time performance feedback, achieving synergistic enhancements in scheduling and modeling.

[0076] This paper deploys the optimized feature compression model and behavioral inference structure onto a blockchain-powered cloud computing platform, establishing a clear call link between model input, predicted output, and on-chain execution logic. Through the triggering mechanism of on-chain smart contracts, this paper enables model results to directly drive on-chain resource allocation and task scheduling, establishing a "prediction-as-execution" business automation process, significantly improving system response speed and policy credibility, and ensuring transparent collaboration and efficient operation of business tasks among multiple participants.

[0077] The present invention establishes a model adaptive closed-loop mechanism based on operational status data. By collecting real-time business response logs, cloud resource scheduling feedback, and behavior prediction deviation information, the system can generate incremental training data and perform micro-batch updates and stability verification without leaving the online deployment environment, thereby ensuring the model's continued adaptability and evolutionary capabilities in complex business environments. This online feedback training model effectively solves the problems of traditional models' strong staticity and poor adaptability, ensuring that the system can maintain intelligent, efficient, and robust processing capabilities when faced with changes in business models, fluctuations in service quality, or abnormal user behavior.

[0078] In summary, this invention effectively improves the comprehensive performance of commercial information systems in data expression, behavior prediction, resource scheduling, and system evolution through the data trust mechanism guaranteed by blockchain, the intelligent modeling capability supported by the deep compression model, the multi-objective-driven structural optimization algorithm, and the continuous learning mechanism guided by operation feedback. It has high innovation, adaptability, and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0080] Figure 1 This is a flowchart of a method for a blockchain-based cloud computing business information system platform proposed by the present invention;

[0081] Figure 2 This is a system flow chart of a blockchain-based cloud computing business information system platform proposed by the present invention;

[0082] Figure 3 This is a data flow diagram of the blockchain-based cloud computing business information system platform proposed by the present invention. DETAILED DESCRIPTION

[0083] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0084] refer to Figure 1-3 , a cloud computing business information system platform based on blockchain, including:

[0085] Data collection and preprocessing module, used to collect and preprocess business data and build a business feature matrix;

[0086] Feature compression modeling module, used to extract spatial locality and sequence features of the business feature matrix based on a multi-layer convolutional structure;

[0087] The loss function optimization module is used to construct the loss function based on the initial latent feature expression and the reconstructed sample;

[0088] The behavior modeling and performance evaluation module is used to calculate the user behavior prediction error based on the potential feature set, and to construct a blockchain-driven multi-objective performance evaluation result by combining the contract execution delay indicator and service quality indicator in the platform;

[0089] Intelligent optimization module, used to execute the seagull optimization algorithm, initialize the individual seagull weights, construct a multi-objective fitness function, simulate the three-stage trajectory behavior, and coordinately optimize the model structure and scheduling strategy parameters;

[0090] The model deployment module is used to deploy the optimized feature compression model to the blockchain-based cloud computing platform;

[0091] The adaptive update module is used to update the commercial feature compression model in real time based on the operating status data.

[0092] This paper utilizes the three-stage convolution trajectory search mechanism from the Seagull optimization algorithm to perform a global search for the combined encoding of feature compression structures and scheduling strategies, improving model compression accuracy and resource allocation efficiency. By constructing a multi-objective performance evaluation space comprised of behavior prediction, contract delay, and service quality, it dynamically adjusts individual initialization weights and search paths, guiding the optimization process to converge toward optimal predictive stability. Furthermore, by integrating blockchain smart contract execution with operational status data feedback, it enables continuous updates and closed-loop optimization after model deployment, enhancing the system's adaptability and intelligence in complex business environments.

[0093] In this embodiment, the modules are connected through the following methods:

[0094] S1. Collect and pre-process business data to construct a business feature matrix;

[0095] S2. Convolutional variational autoencoders are used to build a business feature compression model. The multi-layer convolutional structure extracts the spatial locality and sequence characteristics of the business feature matrix, maps high-dimensional on-chain business data to the latent space, and generates initial latent feature expressions and reconstruction samples.

[0096] S3. Construct a loss function based on the initial latent feature expression and the reconstructed samples, and output a latent feature set for business behavior modeling and resource scheduling.

[0097] S4. Calculate the user behavior prediction error based on the potential feature set, and construct a blockchain-driven multi-objective performance evaluation result by combining the contract execution delay indicator and the service quality indicator;

[0098] S5. Using the Seagull optimization algorithm, we optimized the commercial feature compression model based on the multi-objective performance evaluation results, simulated the three-stage trajectory pattern of the Seagull attack maneuvering behavior, and performed collaborative global search optimization.

[0099] S6. Deploy the optimized business feature compression model in the cloud computing platform operating environment supported by blockchain, drive business automation processing based on behavior prediction results, and use on-chain logic to control behavior reasoning and call resource execution strategies;

[0100] S7. Collect operating status data during operation, and update the business feature compression model in real time based on the operating status data.

[0101] This paper uses a convolutional variational autoencoder to extract the spatial and sequence features of on-chain business data. By constructing a joint loss function, it generates a stable latent feature set, improving the accuracy of behavioral modeling. The Seagull optimization algorithm is introduced to simulate three-stage convolutional trajectories, collaboratively optimizing the compression model structure and scheduling strategy to enhance the system's global responsiveness to multi-objective performance. Combining on-chain execution logic with operational data feedback, the system can dynamically update model parameters, achieving closed-loop optimization of the entire process from modeling and prediction to resource allocation, significantly improving the intelligence and adaptability of business processing flows.

[0102] In this embodiment, commercial business data includes order transaction information, customer behavior logs, smart contract execution records, service access records, marketing activity data, cloud resource usage, and on-chain ledger interaction data.

[0103] This invention integrates order transactions, customer behavior, contract execution, service access, marketing activities, cloud resource usage, and on-chain interactions to construct a full-dimensional feature matrix, enabling behavioral modeling and state perception across the entire business process. By integrating heterogeneous data, the system can accurately extract user behavior patterns and resource usage trends, enhancing the model's adaptability and predictive capabilities in complex business environments. This provides high-quality data support for subsequent intelligent scheduling and automated execution, improving system response speed and policy execution accuracy.

[0104] In this embodiment, the operation status data includes business response logs, cloud resource scheduling feedback information and prediction deviation indicators.

[0105] This method utilizes business response logs, cloud resource scheduling feedback information, and prediction deviation indicators to construct an operational status dataset, enabling full-cycle perception and performance tracking of the system's operational process. Through real-time acquisition and structured processing, this method can dynamically identify model performance fluctuations and resource allocation bottlenecks, driving micro-batch updates and adaptive structural optimization of the feature compression model. Combined with a closed-loop feedback mechanism, the system improves resource utilization while maintaining prediction accuracy, achieving continuous improvements in model stability and business response efficiency, and maintaining efficient and intelligent operation in diverse business scenarios.

[0106] In this embodiment, S2 includes the following specific steps:

[0107] S21. Perform multi-scale convolution processing on the business feature matrix, using multiple convolution kernels to extract the local correlation features of the on-chain business data in the spatial structure and generate a spatial feature representation;

[0108] S22, performing batch normalization and nonlinear activation transformation on the spatial feature representation in sequence to obtain normalized spatial features;

[0109] S23. Perform sliding window modeling on the normalized spatial features in chronological order, extract the temporal change pattern of the on-chain business data, and generate a temporal feature representation;

[0110] S24, inputting the temporal feature representation into the encoder unit of the variational autoencoder, mapping it to the latent space through the probability distribution reparameterization operation, and generating an initial latent feature expression;

[0111] S25, inputting the initial latent feature expression into the decoder unit of the variational autoencoder to generate a reconstructed feature representation;

[0112] S26. Perform structural alignment and shape matching on the reconstructed feature representation and the commercial feature matrix to generate a reconstructed sample.

[0113] This paper extracts the spatial locality and temporal evolution characteristics of on-chain business data through multi-scale convolution processing and sliding window modeling of the business feature matrix, enhancing the depth and stability of feature expression. Combined with the reparameterization mechanism of the variational autoencoder, high-dimensional data is compressed and mapped to the latent space, generating a reconfigurable and statistically consistent feature expression. By outputting reconstructed samples through the decoder and aligning them with the original matrix, the business structure information and semantic integrity are effectively maintained, achieving efficient compression and reversible modeling of complex on-chain business data, and providing a stable feature foundation for subsequent behavior prediction and scheduling optimization.

[0114] In this embodiment, S3 includes the following specific steps:

[0115] S31, aligning the shapes of the initial latent feature expression and the reconstructed samples with the commercial feature matrix to construct a reconstruction comparison tensor set;

[0116] S32, calculating a reconstruction error tensor based on the reconstruction comparison tensor set by element-wise difference;

[0117] S33, modeling the mean and variance of the distribution of the initial latent feature expression in the latent space, and constructing a distribution deviation tensor from the standard multidimensional normal distribution;

[0118] S34. Jointly model the reconstruction error tensor and the distribution deviation tensor to construct a joint loss function:

[0119]

[0120] Among them, L total represents the joint loss function value, Represents the original eigenvalue of the mth dimension of the nth sample in the input commercial feature matrix, Represents the feature value of the mth dimension of the nth sample reconstructed by the decoder, N represents the total number of input samples, and M represents the number of feature dimensions in each sample. represents the square of the mean of the nth sample on the kth latent dimension, represents the square of the standard deviation of the nth sample on the kth potential dimension, K represents the total number of dimensions of the latent space, n represents the input sample index, m represents the input feature dimension index, and k represents the latent space dimension index;

[0121] S35. Apply the business feature compression model to the business feature matrix input, use the encoder to obtain the final latent feature expression, and generate a latent feature set for business behavior modeling and resource scheduling.

[0122] This method constructs a set of reconstruction comparison tensors of initial latent features, reconstructed samples, and the original feature matrix to accurately calculate sample-level reconstruction error. It then combines the mean and variance parameters of the latent space to model structural deviations from the standard normal distribution, forming a joint loss function. This mechanism effectively integrates data fidelity and latent expression standardization to achieve dual-constraint optimization of the compression model. The resulting latent features not only possess strong information restoration capabilities but also possess a stable behavioral modeling structure, providing highly consistent and highly available feature input for business forecasting and resource scheduling.

[0123] In this embodiment, S4 includes the following specific steps:

[0124] S41. Pass the potential feature set as input to the on-chain user behavior inference structure, which includes a feature encoding unit and a time-dependency modeling unit to construct an on-chain behavior trend modeling path;

[0125] S42. Pair the potential feature set with the historical user behavior records on the platform in a time series manner, input the data into the on-chain user behavior inference structure, and output the user behavior prediction result tensor through sequence pattern modeling and evolution trend generation mechanism;

[0126] S43. Compare the user behavior prediction result tensor with the actual on-chain behavior record and calculate the prediction error tensor:

[0127]

[0128] Among them, M predict Prediction consistency metric, N represents the total number of input samples, n represents the input sample index, represents the joint loss function value of the nth sample, Represents the prediction error of the nth sample;

[0129] S44. Extract the smart contract execution records corresponding to the prediction samples from the blockchain platform, calculate the task confirmation delay and transaction settlement delay indicators in the contract process based on the prediction consistency measurement value, and generate a contract execution delay indicator set;

[0130] S45. Combine the platform service operation log to obtain the service call success rate, response time and resource utilization efficiency corresponding to the prediction sample, and generate a set of service quality indicators;

[0131] S46. Normalize and vector-fuse the contract execution delay indicator set and the service quality indicator set to generate a multi-objective performance evaluation result.

[0132] This method effectively captures user behavior evolution patterns by inputting a potential feature set into an on-chain user behavior inference structure and constructing a behavioral trend path using a time-dependent modeling unit. Based on behavioral prediction errors and model structure deviations, a prediction consistency metric is defined, which is then further correlated with smart contract latency and service quality indicators to construct a multi-dimensional performance evaluation vector. This method enables collaborative modeling of behavioral reasoning, contract execution, and service feedback, providing a multi-objective evaluation basis for subsequent optimization strategies, improving the system's responsiveness to dynamic behaviors and the accuracy of scheduling strategies, and enabling more efficient business process decision support.

[0133] In this embodiment, S5 includes the following specific steps:

[0134] S51. Based on the Seagull Algorithm, the Seagull Optimization Algorithm introduces a dynamic initialization mechanism based on behavior prediction consistency measurement and potential distribution structure. It differentially weights the initial population through a prediction stability weighting strategy, constructs the importance weight vector of the individual Seagulls, initializes the Seagull individuals of the Seagull Optimization Algorithm, and sets the initial solution of each Seagull individual as a joint encoding of a set of commercial feature compression model parameters and cloud resource scheduling strategy parameters:

[0135]

[0136] Among them, W (i) represents the initial importance weight of the i-th seagull individual, represents the behavior prediction consistency measure of the nth sample, i represents the seagull individual index, μ n,k represents the mean encoder output of the nth sample on the kth potential dimension, n represents the input sample index, σ n,k represents the standard deviation of the nth sample on the kth potential dimension, μ n,k represents the mean of the nth sample on the kth latent dimension, K represents the total number of dimensions of the latent space, and k represents the dimension index of the latent space;

[0137] S52. Perform fitness evaluation on each individual seagull. After normalizing the three indicators of prediction error, contract delay, and service quality corresponding to each individual seagull, they are weighted and summed using weight coefficients of 0.5, 0.3, and 0.2, respectively, to generate a comprehensive fitness score.

[0138] S53. Based on the position guidance mechanism and nonlinear contraction update mechanism for the seagull with the highest comprehensive fitness score, the trajectories of all seagulls are adjusted to simulate the three-stage pattern of seagull attack and maneuver behavior, including the initial convergent flight stage, the mid-term circling stage, and the final precise attack stage, to generate an updated solution set.

[0139] S54, based on the updated solution set, re-evaluate the fitness value of the seagull individuals and update them according to the global optimal retention mechanism until the maximum number of iterations 100 is met to generate the optimal solution group;

[0140] S55. Decode the optimal solution into commercial feature compression model structure parameters, and output the optimized commercial feature compression model configuration.

[0141] Based on the traditional Seagull algorithm, this invention introduces a dynamic initialization mechanism driven by behavioral prediction consistency and potential distribution structure, constructs a weight vector for the importance of individual Seagulls, and realizes differentiated configuration of individual search capabilities. A multi-objective normalized weighting strategy comprehensively considers prediction error, contract delay, and service quality to improve the accuracy of fitness assessment. A global search process that simulates the three-stage convolution trajectory of the Seagull is adopted to effectively avoid local optimal traps. The maximum number of iterations is set to control the convergence rhythm, ensuring the coordinated optimization of the model structure and scheduling parameters, and enhancing the system's global adaptability and optimization stability in complex business scenarios.

[0142] In this embodiment, S6 includes the following specific steps:

[0143] S61. Deploy the optimized business feature compression model to the feature processing unit in the blockchain-based cloud computing platform to perform real-time access and compression processing of the on-chain business data;

[0144] S62. Based on the potential feature tensor output by the compression model, behavior prediction is performed to generate a set of prediction results for user behavior, resource consumption trends, and contract fulfillment likelihood;

[0145] S63. Based on the set of prediction results, match the preset on-chain policy template and generate a business automation process configuration;

[0146] S64. Write the business automation process configuration into the smart contract trigger mechanism in the blockchain platform, and control the behavior reasoning call path corresponding to the behavior prediction result through the on-chain logic;

[0147] S65. Based on the behavioral reasoning call path, the corresponding resource execution policy configuration is read from the chain, and the relevant computing nodes in the cloud platform are triggered to complete resource allocation, task scheduling and processing response.

[0148] This method deploys the optimized business feature compression model to the blockchain cloud platform feature processing unit, supporting real-time access and compression of on-chain data. Behavior predictions are performed using the potential features output by the model, automatically matching on-chain policy templates to generate business process configurations. Prediction results drive resource reasoning and task execution through a smart contract trigger mechanism, implementing an automated "prediction-as-execution" logic. This method establishes a closed-loop path between models, predictions, and on-chain contract control, significantly improving the intelligence, response efficiency, and execution accuracy of business processing flows and adapting to the needs of complex business dynamic scenarios.

[0149] In this embodiment, S7 includes the following specific steps:

[0150] S71. During the operation of the blockchain-based cloud computing platform, collect operational status data during business execution;

[0151] S72. Perform data cleaning and structural processing on the operating status data to generate a standardized status data set;

[0152] S73. Perform feature alignment processing on the normalized state dataset and the initial business feature matrix to generate an incremental training dataset containing historical features and current feedback;

[0153] S74. Input the incremental training data set into the deployed commercial feature compression model, execute the micro-batch incremental training process, and generate an intermediate updated commercial feature compression model parameter set;

[0154] S75. Perform stability evaluation and performance verification on the intermediate updated commercial feature compression model parameter set to determine whether the preset update conditions are met. If so, replace the current commercial feature compression model parameters.

[0155] S76. Redeploy the replaced business feature compression model to the cloud computing platform operating environment, continue to receive on-chain business data and behavior prediction inputs, and form an adaptive closed-loop update mechanism for the business feature compression model.

[0156] This method collects state data such as response logs, scheduling feedback, and prediction deviations from the blockchain cloud platform's operations to construct a standardized incremental training set and perform micro-batch training to dynamically update the model. Through stability assessment and performance verification, the system can determine and replace current model parameters, ensuring that the compression model continues to adapt to business changes. The updated model is automatically deployed to the operational environment, forming a self-closed-loop learning mechanism that enables real-time evolution of the model structure and behavior prediction capabilities, improving the system's responsiveness and intelligent decision-making in high-frequency trading and complex task scheduling scenarios.

[0157] Example 1:

[0158] To verify the feasibility of this invention, we applied it to a blockchain-based cloud computing business information system deployed by a large cross-border e-commerce platform. This platform processes over 600,000 order transactions daily, serving approximately 380,000 active users. This system involves multi-dimensional business processes, including order payment, service fulfillment, logistics scheduling, and marketing strategies. Business data is recorded in real time on the blockchain ledger. This system exhibits typical characteristics such as high-frequency trading, diverse behavior, dynamic resources, and on-chain execution, making it a suitable testing scenario for this invention.

[0159] During the actual deployment process, the platform's original system had multiple problems: First, the high-dimensional on-chain business data was not effectively compressed before entering the modeling stage, resulting in long training time and unstable results for the behavior prediction model; second, the resource scheduling rules were based on manually formulated static strategies and could not dynamically adjust resource priorities according to user behavior trends, resulting in resource utilization remaining below 62% for a long time; third, the prediction model update cycle was long, with parameter iteration only performed once a month through full retraining, and the system was slow to respond to sudden business changes.

[0160] To address these issues, the platform's technical team deployed the system described in this invention on its existing cloud platform. The system first uses a data acquisition and preprocessing module to collect real-time on-chain data, including order transaction data, user access behavior, service call logs, and smart contract execution records. This data is then standardized to construct a business feature matrix. This feature matrix is ​​then input into a convolutional variational autoencoder model for feature compression. A multi-layer convolutional structure extracts local spatial features and time series behavior patterns, generating low-dimensional latent feature expressions and reconstructed samples. The system extracts mean and standard deviation parameters from the encoder's latent space output and, combined with on-chain behavior records, constructs a joint loss function for training optimization.

[0161] During the behavioral modeling phase, a stable set of latent features is fed into the on-chain behavioral inference structure. Trend predictions are then made based on historical user behavior and contract fulfillment records. A behavioral prediction consistency metric is then calculated based on the deviation between the predicted results and actual behavior. Furthermore, the system extracts metrics such as contract response time, task confirmation latency, and service success rate from the chain to construct a multi-objective performance evaluation vector encompassing behavioral prediction error, contract execution latency, and service quality.

[0162] The model structure and resource scheduling strategy are jointly optimized through the Seagull optimization algorithm. Based on the Seagull three-stage convolution trajectory pattern, the system searches for the optimal solution within a maximum of 100 iterations, and finally selects the optimal compression model structure with a latent space dimension of 32, 4 convolution layers, and a convolution kernel size of 5×5. The scheduling strategy parameters are set with a resource scheduling priority weight of 0.7 and a resource reservation ratio of 0.25.

[0163] The optimized compression model and scheduling strategy are deployed to a blockchain-based cloud platform and linked to the on-chain smart contract module. When a new round of on-chain transactions occurs, the system automatically converts the prediction results into contract inference calls and, through on-chain rules, automatically triggers task resource allocation and service response processes, forming a closed-loop prediction-driven automated execution mechanism.

[0164] Table 1 Comparison of optimization effects of a cloud computing business information system platform based on blockchain

[0165]

[0166] Table 1 demonstrates the superior performance of the proposed system in terms of order processing efficiency, model compression capabilities, behavior prediction accuracy, and resource scheduling response speed, significantly outperforming traditional systems. The model, after latent feature compression, not only reduces feature redundancy by over 90%, but also improves behavior prediction accuracy by over 20%, providing a more precise basis for contract scheduling. Resource utilization increased by over 20%, and service response time and scheduling decision time decreased significantly, making the overall system more intelligent, efficient, and stable.

[0167] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A cloud computing business information system platform based on blockchain, characterized by: include: Data collection and preprocessing module, used to collect and preprocess business data and build a business feature matrix; Feature compression modeling module, used to extract spatial locality and sequence features of the business feature matrix based on a multi-layer convolutional structure; The loss function optimization module is used to construct the loss function based on the initial latent feature expression and the reconstructed sample; The behavior modeling and performance evaluation module is used to calculate the user behavior prediction error based on the potential feature set, and to construct a blockchain-driven multi-objective performance evaluation result by combining the contract execution delay indicator and service quality indicator in the platform; Intelligent optimization module, used to execute the seagull optimization algorithm, initialize the individual seagull weights, construct a multi-objective fitness function, simulate the three-stage trajectory behavior, and coordinately optimize the model structure and scheduling strategy parameters; The model deployment module is used to deploy the optimized feature compression model to the blockchain-based cloud computing platform; The adaptive update module is used to update the commercial feature compression model in real time based on the operating status data.

2. A cloud computing business information system platform based on blockchain according to claim 1, characterized in that: The modules are implemented as follows: S1. Collect and pre-process business data to construct a business feature matrix; S2. Convolutional variational autoencoders are used to build a business feature compression model. The multi-layer convolutional structure extracts the spatial locality and sequence characteristics of the business feature matrix, maps high-dimensional on-chain business data to the latent space, and generates initial latent feature expressions and reconstruction samples. S3. Construct a loss function based on the initial latent feature expression and the reconstructed samples, and output a latent feature set for business behavior modeling and resource scheduling. S4. Calculate the user behavior prediction error based on the potential feature set, and construct a blockchain-driven multi-objective performance evaluation result by combining the contract execution delay indicator and the service quality indicator; S5. Using the Seagull optimization algorithm, we optimized the commercial feature compression model based on the multi-objective performance evaluation results, simulated the three-stage trajectory pattern of the Seagull attack maneuvering behavior, and performed collaborative global search optimization. S6. Deploy the optimized business feature compression model in the cloud computing platform operating environment supported by blockchain, drive business automation processing based on behavior prediction results, and use on-chain logic to control behavior reasoning and call resource execution strategies; S7. Collect operating status data during operation, and update the business feature compression model in real time based on the operating status data.

3. A blockchain-based cloud computing business information system platform according to claim 2, characterized in that: The commercial business data includes order transaction information, customer behavior logs, smart contract execution records, service access records, marketing activity data, cloud resource usage and on-chain ledger interaction data.

4. A cloud computing business information system platform based on blockchain according to claim 2, characterized in that: The operating status data includes business response logs, cloud resource scheduling feedback information and prediction deviation indicators.

5. The cloud computing business information system platform based on blockchain according to claim 2 is characterized in that: The S2 includes the following specific steps: S21. Perform multi-scale convolution processing on the business feature matrix, using multiple convolution kernels to extract the local correlation features of the on-chain business data in the spatial structure and generate a spatial feature representation; S22, performing batch normalization and nonlinear activation transformation on the spatial feature representation in sequence to obtain normalized spatial features; S23. Perform sliding window modeling on the normalized spatial features in chronological order, extract the temporal change pattern of the on-chain business data, and generate a temporal feature representation; S24, inputting the temporal feature representation into the encoder unit of the variational autoencoder, mapping it to the latent space through the probability distribution reparameterization operation, and generating an initial latent feature expression; S25, inputting the initial latent feature expression into the decoder unit of the variational autoencoder to generate a reconstructed feature representation; S26. Perform structural alignment and shape matching on the reconstructed feature representation and the commercial feature matrix to generate a reconstructed sample.

6. The cloud computing business information system platform based on blockchain according to claim 2 is characterized in that: The S3 includes the following specific steps: S31, aligning the shapes of the initial latent feature expression and the reconstructed samples with the commercial feature matrix to construct a reconstruction comparison tensor set; S32, calculating a reconstruction error tensor based on the reconstruction comparison tensor set by element-wise difference; S33, modeling the mean and variance of the distribution of the initial latent feature expression in the latent space, and constructing a distribution deviation tensor from the standard multidimensional normal distribution; S34. Jointly model the reconstruction error tensor and the distribution deviation tensor to construct a joint loss function: Among them, L total represents the joint loss function value, Represents the original eigenvalue of the mth dimension of the nth sample in the input commercial feature matrix, Represents the feature value of the mth dimension of the nth sample reconstructed by the decoder, N represents the total number of input samples, and M represents the number of feature dimensions in each sample. represents the square of the mean of the nth sample on the kth latent dimension, represents the square of the standard deviation of the nth sample on the kth potential dimension, K represents the total number of dimensions of the latent space, n represents the input sample index, m represents the input feature dimension index, and k represents the latent space dimension index; S35. Apply the business feature compression model to the business feature matrix input, use the encoder to obtain the final latent feature expression, and generate a latent feature set for business behavior modeling and resource scheduling.

7. The cloud computing business information system platform based on blockchain according to claim 2 is characterized in that: The S4 includes the following specific steps: S41. Pass the potential feature set as input to the on-chain user behavior inference structure, which includes a feature encoding unit and a time-dependency modeling unit to construct an on-chain behavior trend modeling path; S42. Pair the potential feature set with the historical user behavior records on the platform in a time series manner, input the data into the on-chain user behavior inference structure, and output the user behavior prediction result tensor through sequence pattern modeling and evolution trend generation mechanism; S43. Compare the user behavior prediction result tensor with the actual on-chain behavior record and calculate the prediction error tensor: Among them, M predict Prediction consistency metric, N represents the total number of input samples, n represents the input sample index, represents the joint loss function value of the nth sample, Represents the prediction error of the nth sample; S44. Extract the smart contract execution records corresponding to the prediction samples from the blockchain platform, calculate the task confirmation delay and transaction settlement delay indicators in the contract process based on the prediction consistency measurement value, and generate a contract execution delay indicator set; S45. Combine the platform service operation log to obtain the service call success rate, response time and resource utilization efficiency corresponding to the prediction sample, and generate a set of service quality indicators; S46. Normalize and vector-fuse the contract execution delay indicator set and the service quality indicator set to generate a multi-objective performance evaluation result.

8. The cloud computing business information system platform based on blockchain according to claim 2 is characterized in that: The S5 includes the following specific steps: S51. Based on the Seagull Algorithm, the Seagull Optimization Algorithm introduces a dynamic initialization mechanism based on behavior prediction consistency measurement and potential distribution structure. It differentially weights the initial population through a prediction stability weighting strategy, constructs the importance weight vector of the individual Seagulls, initializes the Seagull individuals of the Seagull Optimization Algorithm, and sets the initial solution of each Seagull individual as a joint encoding of a set of commercial feature compression model parameters and cloud resource scheduling strategy parameters: Among them, W (i) represents the initial importance weight of the i-th seagull individual, represents the behavior prediction consistency measure of the nth sample, i represents the seagull individual index, μ n,k represents the mean encoder output of the nth sample on the kth potential dimension, n represents the input sample index, σ n,k represents the standard deviation of the nth sample on the kth potential dimension, μ n,k represents the mean of the nth sample on the kth latent dimension, K represents the total number of dimensions of the latent space, and k represents the dimension index of the latent space; S52. Perform fitness evaluation on each individual seagull. After normalizing the three indicators of prediction error, contract delay, and service quality corresponding to each individual seagull, they are weighted and summed using weight coefficients of 0.5, 0.3, and 0.2, respectively, to generate a comprehensive fitness score. S53. Based on the position guidance mechanism and nonlinear contraction update mechanism for the seagull with the highest comprehensive fitness score, the trajectories of all seagulls are adjusted to simulate the three-stage pattern of seagull attack and maneuver behavior, including the initial convergent flight stage, the mid-term circling stage, and the final precise attack stage, to generate an updated solution set. S54, based on the updated solution set, re-evaluate the fitness value of the seagull individuals and update them according to the global optimal retention mechanism until the maximum number of iterations 100 is met to generate the optimal solution group; S55. Decode the optimal solution into commercial feature compression model structure parameters, and output the optimized commercial feature compression model configuration.

9. The cloud computing business information system platform based on blockchain according to claim 2 is characterized in that: The S6 comprises the following specific steps: S61. Deploy the optimized business feature compression model to the feature processing unit in the blockchain-based cloud computing platform to perform real-time access and compression processing of the on-chain business data; S62. Based on the potential feature tensor output by the compression model, behavior prediction is performed to generate a set of prediction results for user behavior, resource consumption trends, and contract fulfillment likelihood; S63. Based on the set of prediction results, match the preset on-chain policy template and generate a business automation process configuration; S64. Write the business automation process configuration into the smart contract trigger mechanism in the blockchain platform, and control the behavior reasoning call path corresponding to the behavior prediction result through the on-chain logic; S65. Based on the behavioral reasoning call path, the corresponding resource execution policy configuration is read from the chain, and the relevant computing nodes in the cloud platform are triggered to complete resource allocation, task scheduling and processing response.

10. The cloud computing business information system platform based on blockchain according to claim 2, characterized in that: The S7 includes the following specific steps: S71. During the operation of the blockchain-based cloud computing platform, collect operational status data during business execution; S72. Perform data cleaning and structural processing on the operating status data to generate a standardized status data set; S73. Perform feature alignment processing on the normalized state dataset and the initial business feature matrix to generate an incremental training dataset containing historical features and current feedback; S74. Input the incremental training data set into the deployed commercial feature compression model, execute the micro-batch incremental training process, and generate an intermediate updated commercial feature compression model parameter set; S75. Perform stability evaluation and performance verification on the intermediate updated commercial feature compression model parameter set to determine whether the preset update conditions are met. If so, replace the current commercial feature compression model parameters. S76. Redeploy the replaced business feature compression model to the cloud computing platform operating environment, continue to receive on-chain business data and behavior prediction inputs, and form an adaptive closed-loop update mechanism for the business feature compression model.

Images