Cloud computing-based sci-tech information service method and system

Abstract

The application relates to the technical field of information service, in particular to a science and technology information service method and system based on cloud computing. The method comprises the following steps: collecting multi-source heterogeneous real-time data streams through a distributed terminal sensor network, performing cross-modal semantic feature fusion, and generating a standardized service request feature set; obtaining a historical science and technology information service case library, performing deep association mining and mode refining, simultaneously performing dynamic priority evaluation and resource adaptability analysis, obtaining a service request processing sequence and a cloud resource dynamic scheduling scheme; performing strategy matching, online dynamic evolution and optimization based on the service request processing sequence, generating a science and technology information recommendation strategy and real-time efficiency evaluation parameters; performing encryption encapsulation and intelligent contract binding through a blockchain consensus network, generating a trusted service strategy transaction block, and synchronously updating to the historical science and technology information service case library. The application can greatly improve the service response efficiency of biological science and technology information.

Description

Technical Field

[0001] The present invention relates to the technical field of information service, and in particular to a cloud computing-based scientific and technological information service method and system. Background Art

[0002] The biotechnology industry is a high-tech industry relying on cutting-edge scientific research exploration, massive data support, and rapid technology iteration, covering many subdivided fields such as gene sequencing, biomedical research and development, biological breeding, microbiological research, clinical experiment analysis, and biotechnology achievement transformation. In the whole process of biotechnology scientific research and industrial implementation, massive multi-source heterogeneous scientific and technological information will be continuously generated, including various resources such as scientific research documents, patent technologies, experimental data, research and development schemes, industry standards, clinical cases, and industrial policies. Researchers, research and development institutions and enterprises require high-frequency, high-precision, real-time scientific and technological information push, matching and consulting services to support core work such as scientific research project establishment, experimental optimization, technical breakthrough, and achievement transformation. Therefore, an intelligent, dynamic, trusted, and highly adaptable scientific and technological information service system is the core foundation and key guarantee for reducing the trial-and-error cost of biological scientific research, improving research and development efficiency, and accelerating the innovation and upgrading of the biotechnology industry.

[0003] At present, traditional information service systems can only realize simple reading and retrieval of a single data format, cannot collect service request data streams in a whole domain through distributed sensing, do not have the capability of cross-modal semantic fusion, unified modeling of spatio-temporal attributes and semantic logic, and are difficult to generate standardized and structured service request feature sets, which greatly reduces the response efficiency and stability of biotechnology information services. Summary of the Invention

[0004] Based on this, it is necessary for the present invention to provide a cloud computing-based scientific and technological information service method and system to solve at least one of the above technical problems.

[0005] To achieve the above objective, a cloud computing-based scientific and technological information service method includes the following steps: Step S1: collecting multi-source heterogeneous real-time data streams corresponding to scientific and technological information service requests through a distributed terminal sensor network, performing cross-modal semantic feature fusion on the multi-source heterogeneous real-time data streams, and generating a standardized service request feature set containing spatio-temporal attributes and semantic logic; acquiring a historical scientific and technological information service case library, performing deep association mining and pattern extraction on historical service data in the historical scientific and technological information service case library, and generating a standardized scientific and technological information service strategy feature template library; Step S2: Perform dynamic priority evaluation and resource adaptability analysis on the standardized service request feature set to obtain the service request processing sequence and cloud resource dynamic scheduling scheme; based on the service request processing sequence, perform strategy matching from the standardized science and technology information service strategy feature template library to generate a preliminary science and technology information recommendation strategy set; Step S3: Based on the cloud resource dynamic scheduling scheme, a virtualized strategy execution environment is created on the cloud computing platform. The preliminary set of science and technology information recommendation strategies is deployed in this environment, and real-time feedback data streams are introduced to dynamically evolve and optimize the preliminary set of science and technology information recommendation strategies online, generating science and technology information recommendation strategies and their real-time performance evaluation parameters. Step S4: Through the blockchain consensus network, the technology information recommendation strategy and its real-time performance evaluation parameters are encrypted, encapsulated, and bound to a smart contract to generate a trusted service strategy transaction block, which is then distributed and synchronized to each service node. Based on the trusted service strategy transaction block, the smart contract is automatically executed to feed back and present the technology information service results to the requesting terminal. At the same time, the application performance data of the technology information service results is collected and updated to the historical technology information service case library.

[0006] Furthermore, generating the standardized service request feature set in step S1 includes the following steps: Step S11: Collect multi-source heterogeneous real-time data streams, including text, images, and streaming data, through the terminal sensor network, and extract the spatiotemporal stamps, data source identifiers, and original feature vectors contained in each multi-source heterogeneous real-time data stream; Step S12: Perform modal recognition on the original feature vectors corresponding to each multi-source heterogeneous real-time data stream to obtain structured text features, unstructured image features, and serialized streaming features; Step S13: Based on the structured text features, extract the entity, attribute and relation triples contained in the semantic dependency parsing tree to generate a text semantic network; Step S14: Based on the text semantic network, perform cross-modal semantic anchoring on unstructured image features, extract visual concepts and spatial relationships corresponding to entities in the text semantic network from the image features, and generate a semantically enhanced feature map with image-text alignment. Step S15: Based on the semantic enhancement feature map, perform temporal semantic injection on the serialized streaming features, and perform spatiotemporal correlation modeling between the dynamic change patterns contained in the streaming features and the static concepts in the semantic enhancement feature map to generate a cross-modal semantic feature fusion map. Step S16: Extract subgraph structures containing core semantic concept nodes, inter-concept association paths, and dynamic change patterns from the cross-modal semantic feature fusion graph. Based on the topological complexity and semantic density of the subgraph structures, perform feature reduction and vectorization encoding to generate a standardized service request feature set.

[0007] Furthermore, step S2 includes the following steps: Step S21: Extract the semantic urgency level, resource demand characteristics, and historical processing cycle of similar requests from the standardized service request feature set to generate a multi-dimensional evaluation initial vector; based on the multi-dimensional evaluation initial vector and combined with real-time load monitoring data from the cloud computing platform, calculate the success probability of resource preemption and expected waiting time for each service request through a resource competition game model to generate a request resource competition situation matrix. Step S22: Based on the request resource contention matrix, calculate the comprehensive processing priority score of each service request in the current cloud environment using a dynamic priority adjustment algorithm, and generate a dynamic priority list of service requests; Step S23: Based on the dynamic priority list of service requests, and combined with the real-time availability topology of virtualized computing instances, storage volumes and network bandwidth in the cloud computing platform, perform bipartite graph matching of resource demand and supply to generate a preliminary mapping relationship between service requests and cloud resources. Step S24: Based on the initial mapping relationship, iterative optimization is carried out by introducing resource fragment integration and load balancing constraints to generate a service request processing sequence and a cloud resource dynamic scheduling scheme. This scheme clarifies the processing order of each service request, the resource allocation map, and the expected start timestamp. Step S25: Based on the service request processing sequence, perform strategy matching from the standardized science and technology information service strategy feature template library to generate a preliminary set of science and technology information recommendation strategies.

[0008] Furthermore, step S25 includes the following steps: Extract the standardized service request feature set corresponding to the current service request to be processed from the service request processing sequence, parse the semantic concept nodes and dynamic change patterns contained therein, and generate the semantic intent network of the service request. Based on the semantic intent network of service requests, the standardized science and technology information service strategy feature template library is traversed, and the topological structure matching degree between the semantic intent network and each strategy feature template is calculated by the subgraph isomorphic matching algorithm to generate a preliminary matching degree set. Based on the initial matching degree set, filter the strategy feature templates whose matching degree exceeds the dynamic threshold, obtain their corresponding historical execution trajectory data, and extract the dynamic evolution path and performance inflection point features of the strategy in the trajectory. Based on the dynamic evolution path and performance inflection point characteristics of the strategy, a strategy fitness prediction model is constructed. The semantic intent network of the current service request is input into the model to predict the potential performance decay curve of each candidate strategy when dealing with the current request. By integrating the initial matching degree and potential performance decay curve, a strategy recommendation sequence is generated through multi-objective optimization ranking. The top-ranked candidate strategy logic and parameters in the strategy recommendation sequence are then packaged to generate a preliminary set of science and technology information recommendation strategies.

[0009] Furthermore, predicting the potential performance degradation curve of each candidate strategy in response to the current request includes the following steps: Extract the state transition sequence of the strategy under different service contexts from the dynamic evolution path of the strategy, and identify the critical conditions for the sudden change of strategy behavior from the performance inflection point characteristics to generate the evolution-mutation correlation map of the strategy. Based on the evolution-mutation correlation map, a nonlinear dynamic system model of policy behavior is constructed. This model takes service context features as input and the rate of change of policy internal parameters as system state variables. Extract the temporal evolution trend and logical conflict points of semantic concepts from the semantic intent network of the current service request, and quantify them into a dynamic perturbation vector of the service context; By inputting dynamic disturbance vectors into a nonlinear dynamic system model and solving the Lyapunov exponent spectrum of the model, the stability of the system under disturbances can be analyzed, and the critical points at which the policy behavior will lead to bifurcation or chaos can be predicted. Based on the critical point of system bifurcation or chaos, and combined with the historical basic performance of the strategy, the continuous trajectory of the strategy performance declining as the service request processing time progresses is deduced, and the potential performance decay curve of each candidate strategy is generated.

[0010] Furthermore, step S3 includes the following steps: Step S31: According to the cloud resource dynamic scheduling scheme, on the designated physical cluster of the cloud computing platform, dynamically instantiate multiple lightweight containers according to the resource allocation map to build a virtualization strategy execution environment; Step S32: Extract the logical execution diagram and parameter configuration template corresponding to each strategy from the preliminary set of science and technology information recommendation strategies, load them into the corresponding container in the virtualized strategy execution environment, and generate a set of strategy execution instances; Step S33: Introduce the real-time feedback data stream from the requesting terminal, convert it into an incremental feature vector that the policy execution instance can perceive, and establish a real-time correlation channel between the incremental feature vector and the internal state of each policy execution instance. Step S34: Based on the incremental feature vector, calculate in parallel the deviation between the real-time output of each policy execution instance and the expected target, and generate a dynamic performance evaluation sequence of policy instances; Step S35: Based on the dynamic performance evaluation sequence, start the policy evolution engine in the virtualized policy execution environment, and use genetic operations based on crossover and mutation to adjust and recombine the policy logic and parameters in the policy execution instance set online to generate a new generation of policy population; Step S36: Conduct fitness stress tests on the new generation of strategy population, select strategy execution instances that meet the preset performance threshold, solidify and output their strategy logic and parameters, and generate science and technology information recommendation strategies and their real-time performance evaluation parameters.

[0011] Furthermore, step S34 includes the following steps: Step S341: Obtain the real-time output data stream of each strategy execution instance, extract key indicators representing the accuracy, relevance, and timeliness of scientific and technological information from the output data stream, and generate the real-time output feature vector of the strategy instance. Step S342: Parse the implicit satisfaction signal and behavior correction data of the requesting terminal from the real-time feedback data stream, and generate the dynamic feature vector of the expected target; Step S343: Based on the real-time output feature vector and the dynamic feature vector of the desired target, construct a multi-target bias metric space, and calculate the Euclidean distance and cosine similarity of the output of each strategy execution instance and the dynamic desired target in different dimensions within this space; Step S344: Based on Euclidean distance and cosine similarity of the included angle, calculate the comprehensive deviation score of each policy execution instance in multiple dimensions through a preset weighted fusion rule; Step S345: Perform moving average and trend difference processing on the comprehensive deviation score according to the time window to remove instantaneous noise interference and generate a dynamic performance evaluation sequence of the strategy instance.

[0012] Furthermore, step S4, which involves encrypting and encapsulating the technology information recommendation strategy and its real-time performance evaluation parameters and binding them to a smart contract, includes the following steps: Acquire the science and technology information recommendation strategy and its real-time performance evaluation parameters, encapsulate them with the strategy generation timestamp and execution environment hash value to generate the original strategy data block; The encryption algorithms of multiple consensus nodes in the blockchain network are invoked to perform multi-signature and hash operations on the original policy data block, generating tamper-proof data fingerprints and node consensus proofs; Based on data fingerprints and node consensus proofs, a smart contract code segment containing policy content, fingerprints, proofs, and triggering conditions is constructed, and this smart contract code segment is bound to the original policy data block to generate a smart contract transaction to be confirmed. The smart contract transactions to be confirmed are broadcast to the blockchain network. After consensus is reached through the proof-of-work mechanism, they are packaged and linked to the blockchain to generate a trusted service strategy transaction block. Based on the distributed ledger characteristics of blockchain, trusted service strategy transaction blocks are synchronized to all service nodes. After each node verifies the validity of the block, it triggers the automatic execution logic of the smart contract to complete the service feedback to the requesting terminal.

[0013] Furthermore, after synchronizing the trusted service strategy transaction block to all service nodes, the process also includes: After receiving the trusted service strategy transaction block, the service node parses the block header information and verifies the link relationship between its hash value and the previous block, as well as the validity of the multi-signature. After the block verification is successful, the node virtual machine loads and executes the smart contract code segment, first decrypting and extracting the technology information recommendation strategy from the encrypted original policy data block; Based on the strategy, the smart contract automatically generates a service response message that conforms to the interface specification of the requesting terminal, and pushes the service response message to the corresponding requesting terminal through the secure channel of the blockchain network. The requesting terminal confirms receipt of the service response within a preset time and generates a confirmation receipt containing a receipt timestamp, then broadcasts the confirmation receipt back to the blockchain network as a new transaction. After monitoring the confirmation receipt, the smart contract extracts the receiving timestamp and service policy identifier from the receipt to generate a service completion evidence tag. Based on the service completion evidence tag, it retrieves the interaction logs corresponding to the entire service process from the blockchain's temporary storage area, including on-chain records of policy generation, block confirmation, and policy push. The integrated service completes evidence labeling and interaction logs, and constructs a traceable event sequence for the entire lifecycle of this service by aligning timelines and associating events. Features are extracted from the traceable event sequence to obtain service time consumption, number of interaction rounds, and terminal response latency feature parameters. Combined with the technology information recommendation strategy and its real-time performance evaluation parameters, a multi-dimensional performance profile of the service is generated. The service multi-dimensional performance profile, technology information recommendation strategy, service response message and confirmation receipt are structured, recombined and compressed and encrypted to generate a complete service case data package, and its hash value is used as a unique index and stored in the on-chain storage area of ​​the historical technology information service case library.

[0014] Furthermore, the present invention also provides a cloud computing-based scientific and technological information service system, including a processor, a memory, and a computer program stored in the memory and executable on the processor, for performing the cloud computing-based scientific and technological information service method described above.

[0015] The beneficial effects of this invention are: Compared with the prior art, the beneficial effect of the cloud computing-based scientific and technological information service method proposed by the present invention lies in that it relies on a distributed terminal sensor network to collect multi-source heterogeneous real-time data streams of scientific and technological information service requests, completes cross-modal semantic feature fusion, and constructs a standardized service request feature set with spatio-temporal attributes and semantic logic. Meanwhile, it conducts deep correlation mining and pattern extraction on historical service case data, and builds a standardized scientific and technological information service strategy feature template library. This step realizes global real-time data collection through the distributed sensor network, covers all categories of heterogeneous information resources in biotechnology, uniformly organizes multi-source data features through cross-modal semantic feature fusion, integrates spatio-temporal attributes and semantic logic, and forms a standardized service request feature set, which solves the problems of messy original data, fragmented dimensions and no unified semantic system. By mining historical cases to extract service rules, a standardized strategy template library is constructed, and mature service logic for multiple scenarios is accumulated, which provides a systematic basis for subsequent request adaptation and strategy matching, and improves the organization degree and basic response capability of biotechnology information service from the source. Secondly, dynamic priority assessment and resource adaptability analysis are carried out on the standardized service request feature set to generate a standardized service request processing sequence and a cloud resource dynamic scheduling scheme. Meanwhile, the content of the strategy template library is matched according to the processing sequence, and an initial scientific and technological information recommendation strategy set is output. This step carries out multi-dimensional assessment based on the standardized request feature set, scientifically divides the processing order of service requests, and guarantees the priority implementation of key demands such as core scientific research, technical research and achievement transformation. It dynamically allocates cloud resources in combination with resource adaptability analysis to realize the balance of resource supply and demand, and avoids the problems of resource waste and congestion. It completes scenario-based strategy matching relying on the standardized template library, replaces the traditional shallow retrieval mode, greatly improves the adaptation degree between information service content and biological scientific research scenarios, and effectively solves the problems of disordered response, low resource utilization efficiency and one-sided content matching in traditional information services. Then, a virtualization strategy execution environment is built according to the cloud resource dynamic scheduling scheme, the initial scientific and technological information recommendation strategy set is deployed, and online dynamic evolution and optimization of the strategy are completed in combination with real-time feedback data streams, so as to generate the final recommendation strategy and corresponding time-efficiency and effectiveness assessment parameters. This step realizes independent and efficient deployment of strategies relying on the cloud virtual environment, ensures the parallel and stable operation of multiple types of service strategies, continuously corrects and optimizes the initial recommendation strategy relying on real-time feedback data streams, breaks the disadvantage of solidified traditional strategies, and makes service content fit real-time scientific research demands and dynamic changes of industries. The synchronously generated time-efficiency and effectiveness assessment parameters can completely present the implementation value and application effect of information services, form a quantifiable and traceable service evaluation system, provide an objective basis for subsequent iterative optimization of service strategies and optimization of resource allocation, and greatly improve the dynamic adaptation capability and professional service level of biotechnology information services.Finally, the recommendation strategy and performance evaluation parameters are encrypted and encapsulated using a blockchain consensus network and bound to smart contracts. This generates a trusted service strategy transaction block and completes distributed node synchronization. Smart contracts drive automatic service execution and result feedback, while simultaneously updating the historical service case library. This step leverages the immutability and distributed synchronization characteristics of blockchain to securely store service strategies and performance data. Smart contracts automate the service process, reducing delays and operational errors caused by manual intervention and ensuring efficient, standardized, and reliable information services throughout the entire process. By collecting service application performance data and updating the historical case library, the system continuously expands high-quality service samples and enriches the strategy template system, forming a closed-loop process encompassing demand perception, strategy generation, implementation, effect feedback, and iterative optimization. This continuously improves the credibility, intelligence, and long-term service capabilities of biotechnology information services. Attached Figure Description

[0016] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart illustrating the steps of the cloud computing-based scientific and technological information service method of the present invention. Figure 2 for Figure 1 A detailed flowchart of step S2. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a cloud computing-based method for providing scientific and technological information services. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of the cloud computing-based scientific and technological information service method of the present invention. In this example, the cloud computing-based scientific and technological information service method includes the following steps: Step S1: Collect multi-source heterogeneous real-time data streams corresponding to science and technology information service requests through a distributed terminal sensor network; perform cross-modal semantic feature fusion on the multi-source heterogeneous real-time data streams to generate a standardized service request feature set containing spatiotemporal attributes and semantic logic; obtain a historical science and technology information service case library; perform deep correlation mining and pattern extraction on the historical service data in the historical science and technology information service case library to generate a standardized science and technology information service strategy feature template library. In this embodiment of the invention, targeting the full range of business scenarios in the biotechnology field, including scientific research data analysis, sample detection and judgment, experimental process optimization, and scientific literature tracing, a terminal sensor network architecture with full-domain distributed deployment is relied upon to continuously collect multi-source heterogeneous real-time data streams corresponding to biotechnology information service requests. These data streams encompass multimodal information such as biological experimental record text, microscopic sample imaging images, experimental environment parameter streaming data, and scientific research retrieval behavior data. A cross-modal semantic feature fusion processing flow is initiated for the collected multi-source heterogeneous real-time data streams. First, modal decomposition and feature normalization of the multimodal data are completed. Then, the spatiotemporal stamp attributes, data source identifiers, and original high-dimensional feature vectors of each type of data stream are extracted. Through a multi-level fusion mechanism of semantic dependency deconstruction, visual concept anchoring, and temporal semantic injection, the semantic association barriers between text, images, and streaming data are broken down. This achieves unified modeling of static semantic concepts and dynamic spatiotemporal patterns, generating a standardized biotechnology service request feature set that simultaneously carries spatiotemporal attributes, domain semantic logic, and experimental dynamic features. The system retrieves a historical biotechnology information service case library from the distributed storage on the blockchain. This case library collects strategy operation data, resource scheduling data, service performance data, and terminal interaction log data from all historical biological research services. It then conducts comprehensive deep correlation mining and time-series pattern extraction on the massive historical service data within the case library. Through a multi-level computational mechanism involving feature association clustering, decomposition of time-series evolution patterns, extraction of performance inflection point features, and summarization of scenario adaptation patterns, the system uncovers the inherent correlation patterns among service request characteristics, resource adaptation rules, and strategy operation performance in biological research scenarios. It extracts standardized scenario adaptation strategy patterns and parameter evolution paradigms, and systematically constructs a standardized science and technology information service strategy feature template library with a unified structure, complete features, and adaptability to all biological research scenarios. This provides standardized template support for subsequent strategy matching and intelligent inference.

[0019] Step S2: Perform dynamic priority evaluation and resource adaptability analysis on the standardized service request feature set to obtain the service request processing sequence and cloud resource dynamic scheduling scheme; based on the service request processing sequence, perform strategy matching from the standardized science and technology information service strategy feature template library to generate a preliminary science and technology information recommendation strategy set; In this embodiment of the invention, based on the standardized biotechnology service request feature set generated in step S1, a multi-dimensional dynamic priority assessment and cloud resource adaptability analysis are initiated. The urgency level of the biological research service scenario, experimental data volume characteristics, computing power, storage, and network resource demand characteristics, and historical processing cycle characteristics of similar services are extracted from the feature set and concatenated to generate a multi-dimensional initial assessment vector. This initial vector, along with real-time load monitoring data from the cloud computing platform, is input into a resource competition game model. Through multi-agent game iterative calculations, the resource preemption status and queuing status of each biotechnology service request are quantified, generating a resource competition situation matrix. Based on this situation matrix and a dynamic priority adjustment algorithm, multi-level influencing factors such as platform load status, resource competition intensity, and research task attributes are superimposed. Iterative calculations yield a comprehensive priority score for each service request, and an ordered biotechnology service request processing sequence is generated according to the score distribution rules. Combining the real-time topology availability status of virtualized computing instances, storage volumes, and network bandwidth on the cloud computing platform, a bipartite graph matching operation of resource demand and supply is performed. Multiple rounds of iterative optimization are completed with resource fragmentation integration and global load balancing constraints, converging to generate a service request processing sequence and a dynamic cloud resource scheduling scheme adapted to the biological research scenario. Based on the well-organized service request processing sequence, the semantic intent features, scene topology features, and dynamic change features of biotechnology service requests within the sequence are deconstructed one by one. The standardized science and technology information service strategy feature template library is traversed through the subgraph isomorphic matching mechanism to quantify the matching correlation between the request semantic network and the template topology structure, select highly adaptable strategy templates, and collect and generate a preliminary science and technology information recommendation strategy set covering multiple biological research scenarios.

[0020] Step S3: Based on the cloud resource dynamic scheduling scheme, a virtualized strategy execution environment is created on the cloud computing platform. The preliminary set of science and technology information recommendation strategies is deployed in this environment, and real-time feedback data streams are introduced to dynamically evolve and optimize the preliminary set of science and technology information recommendation strategies online, generating science and technology information recommendation strategies and their real-time performance evaluation parameters. In this embodiment of the invention, relying on the cloud resource dynamic scheduling scheme output in step S2, the physical cluster nodes of the cloud computing platform adapted to the high-precision computing scenarios of biological research are locked. According to the resource allocation map built into the scheme, multiple lightweight isolated containers are dynamically instantiated to build a virtualized strategy execution environment with independent resources, controllable computing power, and network connectivity, enabling the parallel deployment and independent operation of multiple types of biotechnology recommendation strategies. The logical execution graphs and parameter configuration templates of all strategies within the initial set of scientific and technological information recommendation strategies are loaded one by one into the corresponding container units, completing the initial construction of strategy execution instances and generating a set of strategy execution instances that can be dynamically iteratively updated. Real-time feedback data streams uploaded from biological research terminals are continuously accessed, and the feedback data streams undergo structured feature transformation to generate incremental feature vectors that can be recognized and parsed by strategy instances. A real-time correlation channel is established between the incremental feature vectors and the internal operating status of each strategy instance, realizing real-time linkage between dynamic changes in terminal research needs and strategy operating parameters. Based on the incremental feature vectors, multi-instance parallel deviation calculations are performed, comparing the multi-dimensional differences between the real-time output results of each strategy instance and the expected goals of scientific research services, and quantifying and generating a dynamic performance evaluation sequence for strategy instances. A strategy evolution engine within the virtualized environment is initiated based on a dynamic performance evaluation sequence. Through cross-operation of a genetic algorithm, the fusion and inheritance of superior strategy logic and parameters are achieved. Mutation operations are used to innovate and optimize local strategy parameters, generating a new generation of strategy populations adapted to complex biological research scenarios through multiple iterations. Multi-condition stress testing is conducted on this new generation of strategy populations, simulating complex scenarios such as concurrent research data, fluctuating experimental parameters, and changing cloud loads. High-quality strategy instances that meet performance thresholds are selected, and the strategy logic structure and optimal parameter combinations are solidified. Finally, stable and usable biotechnology information recommendation strategies and corresponding real-time performance evaluation parameters are generated.

[0021] Step S4: Through the blockchain consensus network, the technology information recommendation strategy and its real-time performance evaluation parameters are encrypted, encapsulated, and bound to a smart contract to generate a trusted service strategy transaction block, which is then distributed and synchronized to each service node. Based on the trusted service strategy transaction block, the smart contract is automatically executed to feed back and present the technology information service results to the requesting terminal. At the same time, the application performance data of the technology information service results is collected and updated to the historical technology information service case library.

[0022] In this embodiment of the invention, relying on the encrypted evidence storage mechanism of the blockchain distributed consensus network, the biotechnology information recommendation strategy and corresponding real-time performance evaluation parameters generated in step S3 are collected and structured by combining the strategy generation timestamp and container execution environment hash value to construct a complete original strategy data block. The encrypted computing unit of the blockchain multi-consensus nodes is invoked to perform layered hash iteration and multi-signature processing on the original strategy data block, generating tamper-proof data fingerprints and distributed node consensus proofs. Based on the confirmed data, a smart contract code segment carrying trigger execution rules and anomaly verification logic is constructed, completing the encrypted binding of the code segment and the original data block, generating a smart contract transaction to be confirmed. The transaction is broadcast globally to the blockchain consensus network, and the network-wide consensus verification is completed through the proof-of-work mechanism. A trusted service strategy transaction block is packaged and distributed to all cloud service nodes using the distributed ledger synchronization mechanism. Each service node completes block hash verification, chain relationship verification, and signature validity verification. The smart contract is loaded and executed through the node virtual machine, decrypting and extracting standardized biotechnology recommendation strategies. A scientific research information service response message is generated according to the terminal interface specification and pushed to the biological research terminal through an encrypted secure channel. After the terminal completes message reception verification, it generates a confirmation receipt with a timestamp and uploads it to the blockchain network. The smart contract monitors the receipt data and generates a service completion evidence tag. It collects the entire process of on-chain interaction logs and constructs a traceable event sequence for the entire service lifecycle through timeline alignment and event correlation. It extracts characteristic parameters such as service consumption time, interaction rounds, and response latency within the sequence, integrates them with strategy performance parameters to construct a multi-dimensional service performance profile, and compresses and encrypts the profile data, strategy data, response messages, and receipt data to generate a standardized service case data package. Using a hash value as a unique index, it synchronously updates the historical science and technology information service case database, achieving long-term evidence storage and iterative reuse of biotechnology service data.

[0023] Furthermore, generating the standardized service request feature set in step S1 includes the following steps: Step S11: Collect multi-source heterogeneous real-time data streams, including text, images, and streaming data, through the terminal sensor network, and extract the spatiotemporal stamps, data source identifiers, and original feature vectors contained in each multi-source heterogeneous real-time data stream; Step S12: Perform modal recognition on the original feature vectors corresponding to each multi-source heterogeneous real-time data stream to obtain structured text features, unstructured image features, and serialized streaming features; Step S13: Based on the structured text features, extract the entity, attribute and relation triples contained in the semantic dependency parsing tree to generate a text semantic network; Step S14: Based on the text semantic network, perform cross-modal semantic anchoring on unstructured image features, extract visual concepts and spatial relationships corresponding to entities in the text semantic network from the image features, and generate a semantically enhanced feature map with image-text alignment. Step S15: Based on the semantic enhancement feature map, perform temporal semantic injection on the serialized streaming features, and perform spatiotemporal correlation modeling between the dynamic change patterns contained in the streaming features and the static concepts in the semantic enhancement feature map to generate a cross-modal semantic feature fusion map. Step S16: Extract subgraph structures containing core semantic concept nodes, inter-concept association paths, and dynamic change patterns from the cross-modal semantic feature fusion graph. Based on the topological complexity and semantic density of the subgraph structures, perform feature reduction and vectorization encoding to generate a standardized service request feature set.

[0024] In this embodiment of the invention, targeting the entire business scenario of microbial cultivation, cell component analysis, biological sample detection, experimental environment operation and maintenance, and scientific research data analysis in the biotechnology field, a terminal sensor network architecture with full coverage continuously collects multi-source heterogeneous real-time data streams. The full-domain data stream system comprehensively encompasses biotechnology experimental process record text, microscopic high-magnification experimental imaging images, and time-series streaming monitoring data of experimental environment parameters and biological sample status changes, achieving full-domain data collection and coverage of the entire biological research process. For each type of real-time heterogeneous data stream output by the terminal sensor network, deep analysis and extraction of native underlying information are carried out. High-precision spatiotemporal stamp information of the data generation time, terminal device data source identifier of the data, and original feature vector composed of original high-dimensional pixels and values ​​are analyzed and solidified for each data stream. The original structural attributes, spatiotemporal traceability links, and underlying feature distribution patterns of each type of biotechnology data are completely preserved, eliminating the loss and omission of native feature information. Refined modality classification and recognition operations are performed on all collected raw feature vectors. Based on the differences in feature space dimension distribution, data organization structure, and information update form, the multimodal data is accurately classified and split. The resulting data includes structured biological experimental text features with regular structure and standardized fields, unstructured biological microscopic imaging image features with multi-level pixel texture arrangement, and sequential streaming features of experimental states with continuous temporal iteration and dynamic parameter updates. This enables modality stripping and feature regularization of multimodal heterogeneous data. To address the characteristics of structured biotechnology experimental texts that have been properly organized, a multi-level semantic dependency analysis tree is constructed to perform in-depth text deconstruction. Through a tree-like hierarchical decomposition mechanism, the subject-verb-object-adjective-complement dependency relationships of text sentences are broken down layer by layer. The structured triples, consisting of biological entities, inherent attributes of entities, and interaction relationships of entities, are deeply mined within the text content. These triples cover all core biotechnology elements, including microbial strain classification identifiers, physicochemical parameters of biological samples, standardized experimental operation procedures, experimental temperature and humidity environmental constraints, and sample incubation cycle parameters. The correlation transmission links of all triples are pieced together level by level to build a hierarchical, complete, and logically aligned text semantic network, thus comprehensively solidifying the static semantic association system and domain constraint relationships of various scientific and technological elements within the biological experimental text.Using the constructed text semantic network as the global semantic anchoring benchmark, fine-grained cross-modal semantic alignment operations are performed on the features of unstructured biological microscopic imaging images. The pixel grayscale features, texture distribution features, contour boundary features, and spatial arrangement features within the image are traversed region by region to accurately match various biological entity concepts defined within the text semantic network. This identifies visual entity content such as the morphology of microbial strains, internal cell tissue structures, biological sample slice architecture, and placement of experimental equipment in the image. Standardized biological concept labels, relative spatial positions of entities, and regional association features corresponding to the visual content are extracted. A bidirectional and accurate mapping link between text semantic concepts and visual image features is established to generate a semantically enhanced feature map with a high degree of semantic fit between text and image and clear entity correspondence. Based on the static biological concepts, entity attributes, and domain constraint information carried by the semantically enhanced feature map, frame-by-frame temporal semantic injection processing is carried out on the sequential streaming features corresponding to experimental monitoring. The multi-dimensional dynamic change patterns of experimental cultivation temperature fluctuations, continuous changes in environmental humidity, real-time updates of experimental equipment operating parameters, and iterative updates of biological sample activity and morphological state are comprehensively spatiotemporally associated with the static biological entity concepts, experimental parameter standard definitions, and biological cultivation process constraint rules fixed within the semantically enhanced feature map. A deep fusion and association mechanism between static domain semantic knowledge and dynamic temporal change features is established, and a cross-dimensional feature system that takes into account both spatial semantics and temporal changes is constructed. Finally, a complete cross-modal semantic feature fusion map covering text, images, and temporal streaming data is generated. The established cross-modal semantic feature fusion graph undergoes refined subgraph decomposition and feature selection processing. Effective subgraph structures containing core semantic concept nodes of biotechnology, multi-level inter-concept transmission paths, and dynamic change patterns throughout the experimental process are precisely extracted. Through complex quantification operations on the graph topology and semantic density calculations of nodes, redundant feature nodes, invalid association edges, and duplicate feature paths without actual scientific semantic value are systematically eliminated. Effective feature units carrying core business information and scientific characteristics of biotechnology experiments are fully preserved. Simultaneously, the selected high-dimensional sparse features undergo unified-dimensional vectorization encoding processing to achieve dimensional unification and semantic regularization of multimodal heterogeneous features. Finally, a standardized service request feature set with standard dimensions, semantic completeness, high feature concentration, and compatibility with parallel parsing and iterative computation on cloud computing platforms is generated.

[0025] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart of step S2 is shown below. In this embodiment, step S2 includes the following steps: Step S21: Extract the semantic urgency level, resource demand characteristics, and historical processing cycle of similar requests from the standardized service request feature set to generate a multi-dimensional evaluation initial vector; based on the multi-dimensional evaluation initial vector and combined with real-time load monitoring data from the cloud computing platform, calculate the success probability of resource preemption and expected waiting time for each service request through a resource competition game model to generate a request resource competition situation matrix. Step S22: Based on the request resource contention matrix, calculate the comprehensive processing priority score of each service request in the current cloud environment using a dynamic priority adjustment algorithm, and generate a dynamic priority list of service requests; Step S23: Based on the dynamic priority list of service requests, and combined with the real-time availability topology of virtualized computing instances, storage volumes and network bandwidth in the cloud computing platform, perform bipartite graph matching of resource demand and supply to generate a preliminary mapping relationship between service requests and cloud resources. Step S24: Based on the initial mapping relationship, iterative optimization is carried out by introducing resource fragment integration and load balancing constraints to generate a service request processing sequence and a cloud resource dynamic scheduling scheme. This scheme clarifies the processing order of each service request, the resource allocation map, and the expected start timestamp. Step S25: Based on the service request processing sequence, perform strategy matching from the standardized science and technology information service strategy feature template library to generate a preliminary set of science and technology information recommendation strategies.

[0026] In this embodiment of the invention, after obtaining a set of standardized service request features in the field of biotechnology through multimodal feature fusion and encoding, the cloud computing platform initiates a full-process processing of intelligent resource scheduling and scientific and technological information service strategy generation for biotechnology research scenarios. It comprehensively traverses all feature dimensions within the set of standardized service request features, accurately extracting the semantic urgency level, experimental computational and storage resource usage scale, network transmission resource requirements, and complete processing cycle data of historical similar biotechnology research service requests corresponding to each biotechnology service request. Multiple heterogeneous features are numerically regularized and dimensionally concatenated to generate a multidimensional initial evaluation vector with fixed dimensional arrangement, balanced feature weights, and complete representation of the business attributes and historical handling patterns of a single biotechnology research service request. This provides complete pre-feature support for subsequent resource game theory and priority calculation. The constructed multidimensional evaluation initial vector and the real-time global load status monitoring data collected by the cloud computing platform are synchronously input into the resource competition game model. The model relies on the multi-agent game iterative operation mechanism to synchronously simulate the preemption behavior after multiple groups of biotechnology service requests initiate resource applications in parallel. It quantifies the competitive preemption situation of different scientific research requests for virtualized computing power, distributed storage volume space, and dynamic network bandwidth resources. It derives the success probability value and expected waiting time of resource preemption for each request. It comprehensively integrates the game operation results and resource competition status of all service requests in the platform to generate a request resource competition situation matrix with complete row and column dimensions that covers all requests in the entire domain. Based on the global quantitative numerical characteristics carried by the resource competition situation matrix, a multi-factor weighted iterative calculation is carried out through a dynamic priority adjustment algorithm. This algorithm overlays multi-level influencing factors such as the urgency of biological experimental research tasks, the intensity of current resource competition, the efficiency of handling similar tasks in the past, the overall load pressure status of the cloud computing platform, and the resource availability. Through multiple rounds of iterative correction of weight ratios, the influence ratio of each dimension factor is updated successively. The algorithm accurately iteratively calculates the comprehensive priority score of each biotechnology service request in the current cloud platform operating environment. Based on the ranking rules of the comprehensive score, a dynamic priority list of service requests that is updated in real time and has a clear hierarchical order is generated. Strictly following the processing order of the dynamic priority list of service requests, and combining the real-time occupancy status, remaining available capacity, and node distribution topology of all virtualized computing instances, distributed storage volume resources, and dynamic network bandwidth resources within the cloud computing platform, a bidirectional bipartite graph matching topology is constructed between service request resource demand nodes and cloud platform resource supply nodes. Through global optimal bidirectional matching operations, a one-to-one correspondence between demand nodes and supply nodes is completed. The matching process fully considers resource capacity adaptability and the rationality of resource geographical distribution, generating a preliminary mapping correspondence between biotechnology service requests and various computing power, storage, and network resources of the cloud platform.The initial resource mapping structure formed in the first round of matching is subjected to multiple rounds of iterative optimization. A global resource fragmentation integration mechanism is introduced to collect scattered and idle fragmented resources within the platform, complete the merging, reuse, and unified scheduling of fragmented resources, and eliminate the problem of ineffective resource occupation caused by scattered resources being unable to adapt to scientific research tasks. At the same time, relying on the global load balancing constraint mechanism of the cloud computing platform, the unreasonable allocation state of excessive concentrated occupation of local node resources and long-term idle resources of some nodes is adjusted. After multiple rounds of iterative convergence, a stable resource allocation structure is formed, and finally a standardized and unified service request processing sequence and cloud resource dynamic scheduling scheme are generated. The scheme clearly defines the processing order of each biotechnology service request, the exclusive resource allocation topology map, the precise start timestamp of resource mounting and binding, and the resource release and recycling time nodes. Based on the service request processing sequence that has been iteratively optimized, a full-domain feature matching retrieval is performed on each of the ordered biotechnology service requests within the sequence. This involves retrieving various templates from the standardized science and technology information service strategy feature template library, including data analysis strategies, experimental result organization strategies, industry-leading technology information push strategies, research resource matching and scheduling strategies, and experimental risk warning strategies, all adapted to the biological research scenario. This accurately matches the research business scenario, resource requirements, and experimental type characteristics corresponding to each service request. Highly compatible strategy templates are then selected, and parameter adjustments are made. Finally, a preliminary set of science and technology information recommendation strategies that fully adapts to the current biotechnology research service scenario is output. This completes the entire process of intelligent resource scheduling, order organization, and initial service strategy generation for the biotechnology information service scenario on the cloud computing platform.

[0027] Furthermore, step S25 includes the following steps: Extract the standardized service request feature set corresponding to the current service request to be processed from the service request processing sequence, parse the semantic concept nodes and dynamic change patterns contained therein, and generate the semantic intent network of the service request. Based on the semantic intent network of service requests, the standardized science and technology information service strategy feature template library is traversed, and the topological structure matching degree between the semantic intent network and each strategy feature template is calculated by the subgraph isomorphic matching algorithm to generate a preliminary matching degree set. Based on the initial matching degree set, filter the strategy feature templates whose matching degree exceeds the dynamic threshold, obtain their corresponding historical execution trajectory data, and extract the dynamic evolution path and performance inflection point features of the strategy in the trajectory. Based on the dynamic evolution path and performance inflection point characteristics of the strategy, a strategy fitness prediction model is constructed. The semantic intent network of the current service request is input into the model to predict the potential performance decay curve of each candidate strategy when dealing with the current request. By integrating the initial matching degree and potential performance decay curve, a strategy recommendation sequence is generated through multi-objective optimization ranking. The top-ranked candidate strategy logic and parameters in the strategy recommendation sequence are then packaged to generate a preliminary set of science and technology information recommendation strategies.

[0028] In this embodiment of the invention, after completing the regularization of the biotechnology service request processing sequence, the matching of virtualization resource allocation, and the convergence of the global load balancing scheduling scheme on the cloud computing platform, a full-process optimization operation is initiated for precise matching, dynamic deduction, and performance prediction of service strategies for biological research scenarios. From the biotechnology service request processing sequence that has been prioritized and verified for resource adaptation, single biological research service requests awaiting processing are extracted sequentially. The standardized service request feature set generated in the previous multimodal fusion processing stage of the request is retrieved, and multi-level deep deconstruction is carried out on the text semantic features, image visual features, and time-series streaming dynamic feature fusion data sealed within the feature set. The biotechnology-specific semantic concept nodes, the topological transmission paths between different levels of concepts, the dynamic change patterns of parameters throughout the biological experiment, and the state iteration evolution characteristics of microbial samples and experimental environments are deconstructed layer by layer in the cross-modal feature fusion graph. Based on all the static semantic elements and dynamic temporal elements obtained from the deconstruction, a service request semantic intent network with a complete topological structure, tight semantic association, and coherent temporal features is reconstructed. This network comprehensively covers the core intents of various biotechnology services, including microbial strain detection, cell structure analysis, verification of biological sample physicochemical indicators, control of experimental environment parameters, structured analysis of scientific research data, and tracing and matching of scientific literature information. It fully embodies the scenario attributes and business requirements of current scientific research requests. Using the reconstructed, refined semantic intent network of biotechnology services as a unified matching benchmark, it comprehensively traverses all strategy template units within the standardized scientific information service strategy feature template library aimed at the biological research field. Each template has a fixed, exclusive concept node topology, semantic association rules, dynamic parameter adaptation mechanism, and scenario constraints. Relying on the layer-by-layer comparison operation mechanism of the subgraph isomorphic matching algorithm, it conducts a comprehensive bidirectional comparison between the semantic intent network and each strategy template graph at the node level, edge level, and overall topology level. It aligns the biological concept node categories, node association logical paths, dynamic feature arrangement trends, and topological density distribution patterns of the two types of graphs one by one. It quantifies and statistically analyzes the topological overlap ratio, semantic rule fit, and dynamic feature adaptation accuracy between the graphs, and collects the quantitative matching values ​​corresponding to all strategy templates, forming a complete preliminary matching set covering the entire template library. A hierarchical screening and verification process is carried out on all quantitative results within the initial matching degree set. The dynamic threshold on which the matching process is based is dynamically and adaptively updated according to the real-time computing power load of the cloud computing platform, the structural complexity of the biological experiment scenario, and the scale of scientific research data carried in a single request, to select biotechnology strategy feature templates with matching degree values ​​higher than the real-time threshold standard.By retrieving and storing the full-cycle historical execution trajectory data of high-quality templates, the entire process of template operation under different biological experimental conditions and resource loads is fully traced. The dynamic evolution path of strategy parameters iteratively updates within the trajectory is broken down time-series, accurately locating key time-series nodes where strategy service effectiveness fluctuates, rises positively, or falls negatively. The scenario characteristics and parameter states corresponding to all fluctuation nodes are summarized and solidified into a systematic strategy effectiveness inflection point feature dataset. Relying on the dual data support of the strategy dynamic evolution path and time-series effectiveness inflection point features, a strategy fitness prediction model adapted to complex scenarios of biotechnology information services is built. The model has a built-in multi-channel feature mapping structure and a multi-layer time-series iterative calculation module, which can deeply fit the nonlinear adaptation law of strategy operation parameters and dynamic changes in biological research scenarios. The model inputs all the topological features of the semantic intent network corresponding to the current biotechnology service request, the semantic features of biological concepts, and the dynamic temporal evolution features of the experiment. Through the model's internal temporal iterative deduction mechanism, it fits the adaptation and change patterns of each candidate strategy with the current experimental scenario at each time step, deduces the performance fluctuations of different candidate strategies as the service processing progresses, and outputs a complete temporal potential performance decay trend, generating standardized and quantifiable candidate strategy potential performance decay curves. Simultaneously, it integrates the static topological matching metric values ​​output by subgraph isomorphism operations with the dynamic performance decay curve features obtained from the model deduction, constructing a multi-objective optimization ranking operation system that takes into account static structural adaptability, dynamic scenario stability, and long-term service reliability. Through multi-weight iterative correction, it completes the global optimization ranking of all candidate strategies, generating a biotechnology strategy recommendation sequence with clear hierarchical division and well-defined advantages and disadvantages. The top-ranking, highly adaptable candidate strategies in the recommendation sequence undergo logical sorting, rule verification, and unified parameter encapsulation. This solidifies the biological research data analysis logic, precise science and technology information push logic, dynamic cloud resource adaptation and control parameters, and experimental scenario-specific constraint parameters corresponding to various strategies. By integrating all structured logic and quantitative parameters, a preliminary set of science and technology information recommendation strategies with complete structure, self-consistent logic, and direct scheduling and execution by the cloud computing platform is generated.

[0029] Furthermore, predicting the potential performance degradation curve of each candidate strategy in response to the current request includes the following steps: Extract the state transition sequence of the strategy under different service contexts from the dynamic evolution path of the strategy, and identify the critical conditions for the sudden change of strategy behavior from the performance inflection point characteristics to generate the evolution-mutation correlation map of the strategy. Based on the evolution-mutation correlation map, a nonlinear dynamic system model of policy behavior is constructed. This model takes service context features as input and the rate of change of policy internal parameters as system state variables. Extract the temporal evolution trend and logical conflict points of semantic concepts from the semantic intent network of the current service request, and quantify them into a dynamic perturbation vector of the service context; By inputting dynamic disturbance vectors into a nonlinear dynamic system model and solving the Lyapunov exponent spectrum of the model, the stability of the system under disturbances can be analyzed, and the critical points at which the policy behavior will lead to bifurcation or chaos can be predicted. Based on the critical point of system bifurcation or chaos, and combined with the historical basic performance of the strategy, the continuous trajectory of the strategy performance declining as the service request processing time progresses is deduced, and the potential performance decay curve of each candidate strategy is generated.

[0030] In this embodiment of the invention, during the refined prediction calculation stage of potential performance decay curves for candidate biotechnology strategies, a full-dimensional trajectory decomposition analysis is performed on all screened candidate biotechnology service strategies. The complete parameter state transition sequence of each strategy is extracted under different biological experimental contexts, different cloud computing resource load ranges, different biological sample detection conditions, and different research data scales. The state transition sequence fully records the temporal evolution of core parameters such as data parsing weights, information filtering dimensions, resource scheduling ratios, and service response rhythms within the strategy, completely restoring the strategy's operational state change characteristics throughout the entire time period. Simultaneously, a strategy performance inflection point feature dataset accumulated through full-domain traversal is used to systematically analyze the external contextual critical conditions corresponding to fluctuations in strategy service performance such as sudden increases or decreases, steady-state abrupt changes, and adaptation failures. These critical conditions specifically cover complex scenario characteristics such as a rapid increase in the volume of biological experimental detection data, significant shifts in biological sample physicochemical detection parameters, cloud computing platform computing power and bandwidth resources reaching peak load, experimental duration exceeding the conventional processing range, and the parallel superposition of multiple types of biological experimental tasks. By precisely associating and binding the parameter state transition sequences extracted time-series with the efficiency mutation critical conditions identified scenario-by-scenario, a strategy evolution-mutation correlation map is constructed, possessing both time-series evolution representation capabilities and state mutation representation capabilities. This comprehensively depicts the dynamic operation patterns and state transition mechanisms of biotechnology service strategies in complex and ever-changing research scenarios and cloud resource environments. Based on the topological structure features, time-series evolution features, and mutation critical features carried by the evolution-mutation correlation map, a nonlinear dynamic system model of strategy behavior adapted to complex biotechnology service scenarios is built. The model construction process uses multi-dimensional biological research service context features as external input variables and the real-time change rates of internal strategy regulation parameters, data operation parameters, and scenario adaptation parameters as core system state variables. Through multiple rounds of nonlinear iterative calculations, the continuous evolution trajectory of strategy parameters dynamically changing with the research scenario is continuously fitted, accurately depicting the nonlinear fluctuation characteristics and state transition patterns of the entire strategy operation process. Deep, fine-grained feature extraction operations are performed on the semantic intent network corresponding to the current pending biotechnology service request. The temporal evolution patterns of various biotechnology concept nodes within the network are analyzed layer by layer. The semantic temporal evolution trends corresponding to the iterative updates of microbial sample states, dynamic fluctuations of experimental environment parameters, and iterative changes in core research service needs are quantified. The system screens for various logical conflict points within the network structure, such as logical conflicts of experimental parameters, biological concept adaptation deviations, mismatches of temporal features, and missing semantic associations. All temporal evolution features and global logical conflict features are standardized and quantified, and regularized into a multi-dimensional numerical vector with unified dimensions and balanced weights, forming a dynamic perturbation vector of the service context that fully represents the perturbation characteristics of the current biological research scenario.The standardized dynamic disturbance vector is fully input into the constructed nonlinear dynamic system model. Relying on the high-precision nonlinear iterative operation rules built into the model, the whole domain numerical solution is carried out. The numerical distribution of the internal state variables of the system is updated in each iteration cycle. The Lyapunov exponent spectrum corresponding to the system is continuously calculated. By the positive and negative distribution range and numerical fluctuation amplitude of the exponent spectrum, the overall stable state of the nonlinear dynamic system under the dynamic disturbance of the scientific research scenario is determined. The key intervals in which the strategy operation state deviates and the adaptability decreases are gradually locked, and the critical point of periodic bifurcation fluctuation or global chaotic instability of the strategy behavior is accurately defined. Based strictly on the bifurcation critical points and chaotic instability critical points obtained from the dynamic system solution, and combined with historical basic performance data, multi-scenario adaptation performance data, and cloud resource matching performance data accumulated from the long-term operation of biotechnology service strategies, a continuous time-series interpolation and extrapolation mechanism is adopted to compensate for the data discontinuity defects of discrete trajectories. The dynamic change trajectory of strategy performance values ​​along with the continuous increase of biotechnology research service processing time is extrapolated step by step. Smooth, continuous, clear trend, accurate state correspondence, and fully adapted potential performance decay curves of each candidate strategy are generated, providing a comprehensive, stable, and quantitative performance evaluation basis for subsequent multi-objective strategy optimization ranking and intelligent recommendation output.

[0031] Furthermore, step S3 includes the following steps: Step S31: According to the cloud resource dynamic scheduling scheme, on the designated physical cluster of the cloud computing platform, dynamically instantiate multiple lightweight containers according to the resource allocation map to build a virtualization strategy execution environment; Step S32: Extract the logical execution diagram and parameter configuration template corresponding to each strategy from the preliminary set of science and technology information recommendation strategies, load them into the corresponding container in the virtualized strategy execution environment, and generate a set of strategy execution instances; Step S33: Introduce the real-time feedback data stream from the requesting terminal, convert it into an incremental feature vector that the policy execution instance can perceive, and establish a real-time correlation channel between the incremental feature vector and the internal state of each policy execution instance. Step S34: Based on the incremental feature vector, calculate in parallel the deviation between the real-time output of each policy execution instance and the expected target, and generate a dynamic performance evaluation sequence of policy instances; Step S35: Based on the dynamic performance evaluation sequence, start the policy evolution engine in the virtualized policy execution environment, and use genetic operations based on crossover and mutation to adjust and recombine the policy logic and parameters in the policy execution instance set online to generate a new generation of policy population; Step S36: Conduct fitness stress tests on the new generation of strategy population, select strategy execution instances that meet the preset performance threshold, solidify and output their strategy logic and parameters, and generate science and technology information recommendation strategies and their real-time performance evaluation parameters.

[0032] In this embodiment of the invention, step S3, the overall iterative optimization stage, relies on the cloud resource dynamic scheduling scheme output from the preceding process. It locks onto dedicated physical cluster nodes within the cloud computing platform that are adapted to the high-concurrency, high-precision, and low-latency operation characteristics of biotechnology research services. Strictly adhering to the single-node computing power quota, distributed storage volume occupancy quota, internal and external network bandwidth quota, and multi-task hierarchical scheduling priority rules marked by the scheme's built-in resource allocation map, it completes the fully automated dynamic instantiation and deployment of multiple lightweight isolated containers based on the hardware resources of the selected physical cluster. All container deployment processes strictly match the resource consumption characteristics of core biotechnology businesses such as microbial data analysis, biological sample parameter parsing, intelligent scientific literature push, and real-time experimental condition assessment. Through a container resource isolation mechanism, independent computing, storage, and network resource intervals are divided, completely avoiding resource contention and interference caused by the parallel operation of multiple types of biological research tasks. This establishes a virtualized strategy execution environment with full-domain resource isolation, independent computing power scheduling, node network interconnection, and controllable environmental parameters, providing a standardized and highly stable container operating platform for subsequent parallel operation of multiple recommendation strategies, real-time state iteration, and online evolution optimization. From the preliminary set of science and technology information recommendation strategies converged in the previous steps, the structured decomposition of all biotechnology recommendation strategies is completed. The complete operational architecture of each strategy is extracted, and the logical execution topology and standardized parameter configuration templates upon which the strategy operation depends are fully extracted. The logical execution topology fully records the flow path of raw biological research data screening, the matching and association logic of core biological concepts, the rules for parsing experimental parameter thresholds, and the entire topological flow relationship of science and technology information classification and push. The parameter configuration template fully solidifies the multi-dimensional threshold parameters, feature matching weight parameters, semantic association dimension parameters, and time constraint parameters required for strategy operation. The standardized logical execution topology and dedicated parameter configuration templates obtained from the decomposition are precisely mapped and initialized one by one into the corresponding independent container units within the virtualized strategy execution environment. This completes the construction of the strategy operation logic architecture and the initialization of the parameter system within a single container, generating multiple sets of standardized strategy execution instances with independent computing capabilities, parallel scheduling capabilities, dynamic parameter iteration capabilities, and online reconstruction and update capabilities.The system continuously and uninterruptedly receives the full-volume feedback data stream uploaded in real time from the biotechnology service request terminals. This data stream covers multi-dimensional behavioral data such as researchers' browsing trajectories, content verification results, information retention and discarding behaviors, and service interaction operation records for biological strain analysis reports, sample testing data, scientific literature information, and experimental process guidance content. The system performs standardized and structured transformation and feature purification processing on the messy and disordered original terminal feedback data stream, and organizes it into standardized incremental feature vectors that are identifiable, computable, and iteratively updateable for all strategy execution instances according to the biotechnology service feature dimensions. At the same time, a two-way high-speed real-time correlation channel is established between the incremental feature vectors and the internal logical branches, parameter weight matrices, and running status parameters of each strategy execution instance, realizing millisecond-level linkage updates between the dynamic feedback features of terminal users and the running status of strategies within the container. Based on the real-time refreshed incremental feature vector dataset, a multi-container parallel collaborative computing mechanism in the virtualized environment is initiated to synchronously drive all policy execution instances across the entire domain to carry out real-time inference calculations. The real-time output results of each instance are compared with the objective expected goals of biological research services. Fine-grained deviation quantification calculations are completed from four core dimensions: accuracy of bioinformatics matching, completeness of experimental data parsing, timeliness of service response, and adaptability to scientific research scenarios. Multi-dimensional deviation quantification data of all policy instances are collected, sorted, and generated into a dynamic performance evaluation sequence of policy instances with complete dimensions and time sequence. Based on the differences in the performance levels and adaptation defects of the global strategy as fed back by the dynamic performance evaluation sequence, a global intelligent strategy evolution engine is launched within the virtualized strategy execution environment. Relying on the complete population iteration mechanism of the genetic algorithm, the strategy logic topology and multi-dimensional operating parameter groups within the strategy execution instance set are dynamically adjusted and reorganized online without interruption. Cross-operation completes the cross-instance fusion and inheritance of the core logic branches and optimal parameter weights of high-adaptability and high-quality strategies, realizing the population diffusion of excellent strategy characteristics. Mutation operation carries out small random perturbations and logical branch innovation expansion for the strategy's local weak parameters and adaptation defects, making up for the shortcomings of single strategy in scene adaptation. Through multiple rounds of iteration, a new generation of strategy population with more reasonable topology, more scientific parameter ratio, and stronger adaptability to complex biological research scenarios is continuously generated. A comprehensive fitness stress test is conducted on each generation of strategy population generated through iterative updates. This test proactively simulates extreme and complex operating conditions such as massive concurrent access to biological research data, sudden shifts in core experimental parameters, severe fluctuations in the overall load of the cloud computing platform, and mixed parallel operation of multiple types of biological research tasks. The test verifies the information output efficiency, dynamic adaptation capability, long-term operational stability, and fault tolerance capability of all strategy execution instances within the population. High-quality strategy execution instances that consistently meet the preset efficiency thresholds across multiple dimensions of overall efficiency indicators are rigorously selected. The strategy logic topology, core operating rules, and optimal parameter configuration combinations of high-quality instances are fully solidified. Finally, a standardized, directly deployable biotechnology information recommendation strategy system and corresponding real-time efficiency evaluation parameters are output.

[0033] Furthermore, step S34 includes the following steps: Step S341: Obtain the real-time output data stream of each strategy execution instance, extract key indicators representing the accuracy, relevance, and timeliness of scientific and technological information from the output data stream, and generate the real-time output feature vector of the strategy instance. Step S342: Parse the implicit satisfaction signal and behavior correction data of the requesting terminal from the real-time feedback data stream, and generate the dynamic feature vector of the expected target; Step S343: Based on the real-time output feature vector and the dynamic feature vector of the desired target, construct a multi-target bias metric space, and calculate the Euclidean distance and cosine similarity of the output of each strategy execution instance and the dynamic desired target in different dimensions within this space; Step S344: Based on Euclidean distance and cosine similarity of the included angle, calculate the comprehensive deviation score of each policy execution instance in multiple dimensions through a preset weighted fusion rule; Step S345: Perform moving average and trend difference processing on the comprehensive deviation score according to the time window to remove instantaneous noise interference and generate a dynamic performance evaluation sequence of the strategy instance.

[0034] In this embodiment of the invention, the performance evaluation calculation stage continuously collects the full real-time output data stream of all container strategy execution instances within the virtualization strategy execution environment. The data stream comprehensively covers all service output data, including intelligent push results of cutting-edge biotechnology literature, structured data analysis conclusions of microbial experiments, matching content of physicochemical parameters of biological samples, integrated output of core industry technology information, and guidance information for experimental process optimization. The unstructured raw output data stream is processed by hierarchical and multi-dimensional key feature purification and indicator decomposition. Four types of core quantitative indicators that can comprehensively characterize the accuracy of biotechnology information push, the relevance and matching degree of scientific research content, the timeliness of service response and processing, and the adaptability and fit of experimental scenarios are extracted one by one. The heterogeneous indicators with different dimensions and large differences in numerical range are subjected to global normalization and regularization operations, and all indicators are uniformly mapped to the standard numerical range. The system integrates and constructs a real-time output feature vector of strategy instances with unified dimensions, balanced weights, and orthogonal features, which accurately carries the global output quality and service capability characteristics of a single strategy instance at the current time node. The system performs in-depth and refined analysis and deconstruction of the real-time feedback data stream continuously uploaded by biological research terminals, completing the layered separation of explicit interaction data and implicit behavioral data. It accurately mines multi-dimensional behavioral correction data such as implicit satisfaction preferences, professional content modification tendencies, research focus dimension shifts, proactive elimination of invalid and redundant information, and retention and reuse of effective core information by terminal research users regarding biotechnology content. The fragmented and disordered original terminal behavioral feedback information is then structured, categorized, quantified, and feature-fused, integrating and generating a dynamic feature vector of expected targets that fully matches the real service needs of biological research and can be dynamically updated according to users' research needs. This achieves a structured, numerical, and dynamic concrete expression of fuzzy research service needs. Using real-time output feature vectors and expected target dynamic feature vectors as the core dimensional basis, a four-dimensional structured deviation measurement space adapted to biotechnology service scenarios is constructed. The space precisely corresponds to the dimensions of information accuracy, content relevance, service timeliness, and scenario adaptability. Within the standardized deviation measurement space, refined deviation quantification calculations are performed for each strategy instance. The Euclidean distance and cosine similarity between the output vector of each strategy execution instance and the dynamic expected target vector are calculated respectively. The Euclidean distance accurately represents the overall deviation of the two sets of feature vectors in the absolute coordinate position in the high-dimensional space, while the cosine similarity accurately represents the directional fit between the two sets of feature vectors in terms of semantic feature distribution and demand adaptation trend.Combining the core attributes of high precision and professionalism in biotechnology services, a multi-dimensional weighted fusion rule adapted to specific scenarios is established. Differentiated fixed weight coefficients are configured for Euclidean distance and cosine similarity values ​​of different evaluation dimensions. Through a fusion operation method of weighted summation of multi-dimensional values, the one-sidedness of a single distance measure or a single similarity measure is made up for. The spatial location deviation and semantic direction deviation features are fully integrated, and a comprehensive deviation score covering the entire service evaluation index is calculated for each instance, which accurately reflects the overall deviation between the real-time output content of different strategy instances and the user's dynamic scientific research expectations. A standardized, continuous sliding time window of fixed duration is set up to perform a time-series moving average calculation on multiple sets of comprehensive deviation scores generated iteratively within a continuous time series period. This effectively smooths out invalid noise data in the time series dimension caused by instantaneous data fluctuations, occasional user operation deviations, and instantaneous platform load jitters, completely eliminating the interference of single sudden abnormal data on the overall performance evaluation results. At the same time, a first-order trend difference operation is performed on the smoothed time series score sequence to accurately solve the dynamic change gradient of comprehensive deviation as the service processing time progresses. This accurately captures the three evolution trends of strategy performance: continuous decay, steady increase, and dynamic steady state, and fully preserves the time series performance evolution characteristics during the long-term operation of the strategy. Finally, a dynamic performance evaluation sequence of strategy instances with strong time series continuity, excellent anti-interference performance, and objective and stable evaluation results is generated. This provides accurate, reliable, time-series coherent, and dimensionally comprehensive quantitative evaluation support for subsequent strategy genetic evolution iteration, population selection, and high-quality strategy solidification and optimization within the virtualization environment.

[0035] Furthermore, step S4, which involves encrypting and encapsulating the technology information recommendation strategy and its real-time performance evaluation parameters and binding them to a smart contract, includes the following steps: Acquire the science and technology information recommendation strategy and its real-time performance evaluation parameters, encapsulate them with the strategy generation timestamp and execution environment hash value to generate the original strategy data block; The encryption algorithms of multiple consensus nodes in the blockchain network are invoked to perform multi-signature and hash operations on the original policy data block, generating tamper-proof data fingerprints and node consensus proofs; Based on data fingerprints and node consensus proofs, a smart contract code segment containing policy content, fingerprints, proofs, and triggering conditions is constructed, and this smart contract code segment is bound to the original policy data block to generate a smart contract transaction to be confirmed. The smart contract transactions to be confirmed are broadcast to the blockchain network. After consensus is reached through the proof-of-work mechanism, they are packaged and linked to the blockchain to generate a trusted service strategy transaction block. Based on the distributed ledger characteristics of blockchain, trusted service strategy transaction blocks are synchronized to all service nodes. After each node verifies the validity of the block, it triggers the automatic execution logic of the smart contract to complete the service feedback to the requesting terminal.

[0036] In this embodiment of the invention, a complete set of biotechnology information recommendation strategies, along with corresponding dynamic real-time performance evaluation parameters, is used to collect and integrate all valid business data. Specifically, this includes a complete logical architecture, hierarchical execution branches, multi-dimensional operating parameter groups, and quantitative performance scoring indicators for strategies such as microbial community data analysis, biological sample physicochemical parameter comparison, precise matching of biological research literature, intelligent assessment of experimental conditions, and integration and push of research results. Simultaneously, the global time-series timestamp of the iteration and solidification of the entire strategy system is captured, accurately recording the unique time-series node of strategy formation. Furthermore, the global environment hash verification value generated during the execution of the virtualized container strategy is extracted, completely locking down the resource operating status and environment configuration characteristics during the strategy generation stage. Four types of heterogeneous data units—strategy logic data, performance parameter data, time-series stamp data, and environment hash verification data—are encapsulated in a layered, nested structure. Following fixed field arrangement rules, data regularization, field completion, and structural splicing are performed to generate original strategy data blocks with clear field hierarchy, continuous closed-loop structure, and no missing data. The data blocks completely retain the topological operation logic of the biotechnology service strategy, multi-dimensional parameter configurations, dynamic performance evaluation status, strategy generation time-series trajectory, and container operating environment characteristics. A multi-consensus node collaborative encryption operation mechanism under the distributed architecture of the blockchain network is initiated. The independent encryption operation units of the consensus nodes deployed across the entire network are invoked to perform layered parallel hash iteration operations and segmented multi-signature verification on the formed original strategy data blocks. Each consensus node independently performs hash mapping iterations for the time-series field at the beginning of the data block, the strategy body field in the middle, and the performance parameter field at the end, generating fixed-length hash digests for each field layer by layer. Each node completes a nested binding of its own node identity identifier with the hash digest, generating node-specific signature information. The hierarchical hash digests and exclusive signature information output by multiple nodes are interwoven and verified. Through multiple composite hash iteration operations, a globally unified, tamper-resistant global data fingerprint is generated. Simultaneously, a distributed consensus proof is constructed based on the joint signature results of multiple nodes, completing the distributed network-wide ownership confirmation of biotechnology service strategy data. Based on the structured ownership confirmation data of the global data fingerprint and distributed node consensus proof, the business triggering rules and on-chain execution specifications of biotechnology information services are adapted field by field. A dedicated smart contract code segment structure is constructed layer by layer. The code segment solidifies the automatic execution triggering conditions of biotechnology service strategies, terminal service response output specifications, on-chain data dynamic verification logic, cross-node interaction constraint rules, and exception rollback mechanisms. It fully includes the standardized strategy operation main content, global data fingerprint, distributed node consensus proof, scenario-adapted trigger thresholds, and exception handling logic.The process involves cryptographically binding and associating the smart contract code segment with the original strategy data block, establishing a one-to-one mapping and verification relationship between the code execution logic and the original data content, and forming a pending smart contract transaction with mutually corroborating fields, a closed-loop logical linkage, and encrypted data storage. The completed pending smart contract transaction is then broadcast globally to all consensus nodes in the blockchain network. All consensus nodes simultaneously initiate a proof-of-work iterative computation mechanism, continuously iterating and solving the random number constraint condition at the block header to ensure the overall hash value of the block meets the network-wide unified difficulty threshold, completing cross-node data comparison and consistency consensus verification. Once the network-wide consensus verification is complete, the smart contract transaction data, node consensus proofs, and global data fingerprints are packaged into a new block. Relying on the blockchain's inherent hash chain association mechanism, the hash digest at the beginning of the new block is encrypted and linked to the hash digest at the end of the previous block, constructing an irreversible chain block structure and generating a complete and trustworthy biotechnology service strategy transaction block. Relying on the real-time synchronization mechanism of the blockchain system's distributed ledger, the mature trusted service strategy transaction blocks are pushed in batches to all service nodes across the cloud computing platform. Each service node completes the full synchronization and local solidification of block data, laying a solid foundation of underlying block data support for subsequent automatic execution of on-chain smart contracts, terminal security service feedback, and full-process behavior traceability.

[0037] Furthermore, after synchronizing the trusted service strategy transaction block to all service nodes, the process also includes: After receiving the trusted service strategy transaction block, the service node parses the block header information and verifies the link relationship between its hash value and the previous block, as well as the validity of the multi-signature. After the block verification is successful, the node virtual machine loads and executes the smart contract code segment, first decrypting and extracting the technology information recommendation strategy from the encrypted original policy data block; Based on the strategy, the smart contract automatically generates a service response message that conforms to the interface specification of the requesting terminal, and pushes the service response message to the corresponding requesting terminal through the secure channel of the blockchain network. The requesting terminal confirms receipt of the service response within a preset time and generates a confirmation receipt containing a receipt timestamp, then broadcasts the confirmation receipt back to the blockchain network as a new transaction. After monitoring the confirmation receipt, the smart contract extracts the receiving timestamp and service policy identifier from the receipt to generate a service completion evidence tag. Based on the service completion evidence tag, it retrieves the interaction logs corresponding to the entire service process from the blockchain's temporary storage area, including on-chain records of policy generation, block confirmation, and policy push. The integrated service completes evidence labeling and interaction logs, and constructs a traceable event sequence for the entire lifecycle of this service by aligning timelines and associating events. Features are extracted from the traceable event sequence to obtain service time consumption, number of interaction rounds, and terminal response latency feature parameters. Combined with the technology information recommendation strategy and its real-time performance evaluation parameters, a multi-dimensional performance profile of the service is generated. The service multi-dimensional performance profile, technology information recommendation strategy, service response message and confirmation receipt are structured, recombined and compressed and encrypted to generate a complete service case data package, and its hash value is used as a unique index and stored in the on-chain storage area of ​​the historical technology information service case library.

[0038] In this embodiment of the invention, after the global service nodes complete the synchronization and solidification operation of the trusted service strategy transaction block, a closed-loop processing flow is initiated, including multi-level on-chain block legality verification, automatic loading and execution of smart contracts, terminal service response push, receipt storage and traceability, and service performance profile archiving. After all cloud computing service nodes receive the synchronized trusted service strategy transaction block data, they prioritize initiating block header information parsing operations, disassembling the global time-series information, hierarchical hash digest data, front and back blockchain-style index identifiers, and multi-node signature sets stored within the block layer by layer, and conducting legality verification operations dimension by dimension. This verifies the consistency between the block's own hierarchical hash iteration results and the network-wide storage standards, the chain link integrity between the new block and its historical predecessor blocks, and the matching validity of multi-node multi-signature information. It also investigates block transmission anomalies, data tampering risks, and signature forgery issues, ensuring the original authenticity and chain integrity of the on-chain biotechnology strategy data. After all multi-level verifications of the blocks pass, the built-in virtual machines of each service node automatically complete the loading, initialization, parsing, and runtime environment adaptation of the smart contract code segments. They then call the on-chain symmetric decryption mechanism to perform layered and segmented decryption of the encrypted original strategy data blocks. The encrypted structure is deconstructed in reverse according to the field hierarchy, and the internally sealed biotechnology information recommendation strategy logic, parameter configuration rules, and service execution standards are parsed layer by layer to fully extract the complete strategy execution basis for the automated business push of the smart contract. The smart contract operates based on the standardized biotechnology service strategy logic obtained through decryption. It proactively adapts to the dedicated interface transmission protocol, data formatting, message encapsulation specifications, and interaction adaptation standards of the biological research request terminal. It automatically completes the structured reorganization, content regularization, and field adaptation of multi-dimensional research data, generating a standardized service response message that fully conforms to the terminal's parsing standards. The message accurately covers microbial data analysis conclusions, comparison results of biological sample physicochemical parameters, cutting-edge biological research literature resources, guidance schemes for optimizing biological experimental processes, and research data trend analysis. Through the blockchain network's dedicated encrypted and secure transmission channel, the service response message is delivered to the biological research terminal that initiated the research service request. Within the system's preset fixed response time range, the biological research terminal automatically completes the reception, decryption, and content verification of the encrypted service response message. The terminal's backend program automatically generates a reception timestamp carrying the network-wide unified time sequence standard, integrates the terminal's unique identifier, message verification results, and data integrity verification status to generate a standardized structured service confirmation receipt. The receipt data is encapsulated into a new independent transaction unit and broadcast back to the blockchain network for network-wide consensus and notarization. The smart contract continuously monitors the global transaction broadcast data stream of the blockchain network in real time, dynamically captures service confirmation receipt transactions uploaded by the terminal, accurately extracts the standardized receiving timestamp and unique service strategy identifier code sealed inside the receipt, and combines it with the on-chain rights confirmation record to generate an immutable and permanently traceable service completion evidence label.The smart contract proactively retrieves all log resources from the distributed temporary storage area of ​​the blockchain network, collecting logs related to the strategy iteration generation, virtualization container operation parameters, block packaging and network consensus, smart contract execution, and terminal interaction pushes throughout the entire lifecycle of this biotechnology service. This data is integrated into a time-series-continuous and complete set of on-chain interaction logs. Leveraging a global timeline alignment algorithm and a multi-event association matching mechanism, the service completion evidence tags are sequentially linked and event-by-event-associated with the full set of on-chain interaction logs. This comprehensively outlines the entire business flow logic from the initiation of the biotechnology research request, multi-modal feature fusion, intelligent strategy generation, on-chain encrypted rights confirmation, automatic contract push, to terminal receipt confirmation, constructing a closed-loop, precise, and traceable event sequence for the entire service lifecycle. Multi-dimensional refined feature extraction operations are performed on the established traceable event sequences. Objective operational feature parameters such as the overall service processing time cycle, the total number of interaction rounds between the cloud computing platform and the scientific research terminal, the terminal message reception response latency, and the on-chain consensus time parameters are statistically quantified time-series. Combined with the structural parameters of the biotechnology recommendation strategy itself, dynamic performance evaluation parameters, and scenario adaptation parameters, multi-dimensional heterogeneous data deep fusion modeling processing is carried out to generate a multi-dimensional service performance profile that can comprehensively characterize the operational efficiency, content quality, scenario adaptability, and system stability of biotechnology information services. The service multi-dimensional performance profile data, the complete content of standardized biotechnology recommendation strategies, structured service response message data, and terminal service confirmation receipt data are subjected to multi-level structured reorganization, redundant field removal, core feature regularization, and overall compression and encryption processing to generate a standardized biotechnology service case data package with closed-loop structure, reliable encryption, unique fields, and complete time sequence. The formed complete case data package is subjected to a global hash iteration operation, and the final hash digest is used as a globally unique retrieval index. It is uniformly archived and stored in the dedicated distributed storage area of ​​the on-chain historical science and technology information service case library, realizing long-term evidence preservation, fast retrieval, traceability and review, and model iteration optimization and reuse of biotechnology service cases.

[0039] Furthermore, the present invention also provides a cloud computing-based scientific and technological information service system, including a processor, a memory, and a computer program stored in the memory and executable on the processor, for performing the cloud computing-based scientific and technological information service method described above.

[0040] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for providing scientific and technological information services based on cloud computing, characterized in that, Includes the following steps: Step S1: Collect multi-source heterogeneous real-time data streams corresponding to scientific and technological information service requests through a distributed terminal sensor network, perform cross-modal semantic feature fusion on the multi-source heterogeneous real-time data streams, and generate a standardized service request feature set containing spatiotemporal attributes and semantic logic. Acquire a historical science and technology information service case library, conduct in-depth correlation mining and pattern extraction on the historical service data in the historical science and technology information service case library, and generate a standardized science and technology information service strategy feature template library; Step S2: Perform dynamic priority evaluation and resource adaptability analysis on the standardized service request feature set to obtain the service request processing sequence and cloud resource dynamic scheduling scheme; based on the service request processing sequence, perform strategy matching from the standardized science and technology information service strategy feature template library to generate a preliminary science and technology information recommendation strategy set; Step S3: Based on the cloud resource dynamic scheduling scheme, a virtualized strategy execution environment is created on the cloud computing platform. The preliminary set of science and technology information recommendation strategies is deployed in this environment, and real-time feedback data streams are introduced to dynamically evolve and optimize the preliminary set of science and technology information recommendation strategies online, generating science and technology information recommendation strategies and their real-time performance evaluation parameters. Step S4: Through the blockchain consensus network, the technology information recommendation strategy and its real-time performance evaluation parameters are encrypted, encapsulated, and bound to a smart contract to generate a trusted service strategy transaction block, which is then distributed and synchronized to each service node. Based on the trusted service strategy, the transaction block drives the automatic execution of smart contracts, which feeds back and presents the results of technology information services to the requesting terminal. At the same time, it collects the application effectiveness data of the technology information service results and updates them to the historical technology information service case library.

2. The method for providing scientific and technological information services based on cloud computing according to claim 1, characterized in that, Step S1, generating a standardized service request feature set, includes the following steps: Step S11: Collect multi-source heterogeneous real-time data streams, including text, images, and streaming data, through the terminal sensor network, and extract the spatiotemporal stamps, data source identifiers, and original feature vectors contained in each multi-source heterogeneous real-time data stream; Step S12: Perform modal recognition on the original feature vectors corresponding to each multi-source heterogeneous real-time data stream to obtain structured text features, unstructured image features, and serialized streaming features; Step S13: Based on the structured text features, extract the entity, attribute and relation triples contained in the semantic dependency parsing tree to generate a text semantic network; Step S14: Based on the text semantic network, perform cross-modal semantic anchoring on unstructured image features, extract visual concepts and spatial relationships corresponding to entities in the text semantic network from the image features, and generate a semantically enhanced feature map with image-text alignment. Step S15: Based on the semantic enhancement feature map, perform temporal semantic injection on the serialized streaming features, and perform spatiotemporal correlation modeling between the dynamic change patterns contained in the streaming features and the static concepts in the semantic enhancement feature map to generate a cross-modal semantic feature fusion map. Step S16: Extract subgraph structures containing core semantic concept nodes, inter-concept association paths, and dynamic change patterns from the cross-modal semantic feature fusion graph. Based on the topological complexity and semantic density of the subgraph structures, perform feature reduction and vectorization encoding to generate a standardized service request feature set.

3. The method for providing scientific and technological information services based on cloud computing according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Extract the semantic urgency level, resource demand characteristics, and historical processing cycle of similar requests from the standardized service request feature set to generate a multi-dimensional evaluation initial vector; based on the multi-dimensional evaluation initial vector and combined with real-time load monitoring data from the cloud computing platform, calculate the success probability of resource preemption and expected waiting time for each service request through a resource competition game model to generate a request resource competition situation matrix. Step S22: Based on the request resource contention matrix, calculate the comprehensive processing priority score of each service request in the current cloud environment using a dynamic priority adjustment algorithm, and generate a dynamic priority list of service requests; Step S23: Based on the dynamic priority list of service requests, and combined with the real-time availability topology of virtualized computing instances, storage volumes and network bandwidth in the cloud computing platform, perform bipartite graph matching of resource demand and supply to generate a preliminary mapping relationship between service requests and cloud resources. Step S24: Based on the initial mapping relationship, iterative optimization is carried out by introducing resource fragment integration and load balancing constraints to generate a service request processing sequence and a cloud resource dynamic scheduling scheme. This scheme clarifies the processing order of each service request, the resource allocation map, and the expected start timestamp. Step S25: Based on the service request processing sequence, perform strategy matching from the standardized science and technology information service strategy feature template library to generate a preliminary set of science and technology information recommendation strategies.

4. The cloud computing-based scientific and technological information service method according to claim 3, characterized in that, Step S25 includes the following steps: Extract the standardized service request feature set corresponding to the current service request to be processed from the service request processing sequence, parse the semantic concept nodes and dynamic change patterns contained therein, and generate the semantic intent network of the service request. Based on the semantic intent network of service requests, the standardized science and technology information service strategy feature template library is traversed, and the topological structure matching degree between the semantic intent network and each strategy feature template is calculated by the subgraph isomorphic matching algorithm to generate a preliminary matching degree set. Based on the initial matching degree set, filter the strategy feature templates whose matching degree exceeds the dynamic threshold, obtain their corresponding historical execution trajectory data, and extract the dynamic evolution path and performance inflection point features of the strategy in the trajectory. Based on the dynamic evolution path and performance inflection point characteristics of the strategy, a strategy fitness prediction model is constructed. The semantic intent network of the current service request is input into the model to predict the potential performance decay curve of each candidate strategy when dealing with the current request. By integrating the initial matching degree and potential performance decay curve, a strategy recommendation sequence is generated through multi-objective optimization ranking. The top-ranked candidate strategy logic and parameters in the strategy recommendation sequence are then packaged to generate a preliminary set of science and technology information recommendation strategies.

5. The cloud computing-based scientific and technological information service method according to claim 4, characterized in that, The process of predicting the potential performance degradation curve of each candidate strategy in response to the current request includes the following steps: Extract the state transition sequence of the strategy under different service contexts from the dynamic evolution path of the strategy, and identify the critical conditions for the sudden change of strategy behavior from the performance inflection point characteristics to generate the evolution-mutation correlation map of the strategy. Based on the evolution-mutation correlation map, a nonlinear dynamic system model of policy behavior is constructed. This model takes service context features as input and the rate of change of policy internal parameters as system state variables. Extract the temporal evolution trend and logical conflict points of semantic concepts from the semantic intent network of the current service request, and quantify them into a dynamic perturbation vector of the service context; By inputting dynamic disturbance vectors into a nonlinear dynamic system model and solving the Lyapunov exponent spectrum of the model, the stability of the system under disturbances can be analyzed, and the critical points at which the policy behavior will lead to bifurcation or chaos can be predicted. Based on the critical point of system bifurcation or chaos, and combined with the historical basic performance of the strategy, the continuous trajectory of the strategy performance declining as the service request processing time progresses is deduced, and the potential performance decay curve of each candidate strategy is generated.

6. The method for providing scientific and technological information services based on cloud computing according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: According to the cloud resource dynamic scheduling scheme, on the designated physical cluster of the cloud computing platform, dynamically instantiate multiple lightweight containers according to the resource allocation map to build a virtualization strategy execution environment; Step S32: Extract the logical execution diagram and parameter configuration template corresponding to each strategy from the preliminary set of science and technology information recommendation strategies, load them into the corresponding container in the virtualized strategy execution environment, and generate a set of strategy execution instances; Step S33: Introduce the real-time feedback data stream from the requesting terminal, convert it into an incremental feature vector that the policy execution instance can perceive, and establish a real-time correlation channel between the incremental feature vector and the internal state of each policy execution instance. Step S34: Based on the incremental feature vector, calculate in parallel the deviation between the real-time output of each policy execution instance and the expected target, and generate a dynamic performance evaluation sequence of policy instances; Step S35: Based on the dynamic performance evaluation sequence, start the policy evolution engine in the virtualized policy execution environment, and use genetic operations based on crossover and mutation to adjust and recombine the policy logic and parameters in the policy execution instance set online to generate a new generation of policy population; Step S36: Conduct fitness stress tests on the new generation of strategy population, select strategy execution instances that meet the preset performance threshold, solidify and output their strategy logic and parameters, and generate science and technology information recommendation strategies and their real-time performance evaluation parameters.

7. The cloud computing-based scientific and technological information service method according to claim 6, characterized in that, Step S34 includes the following steps: Step S341: Obtain the real-time output data stream of each strategy execution instance, extract key indicators representing the accuracy, relevance, and timeliness of scientific and technological information from the output data stream, and generate the real-time output feature vector of the strategy instance. Step S342: Parse the implicit satisfaction signal and behavior correction data of the requesting terminal from the real-time feedback data stream, and generate the dynamic feature vector of the expected target; Step S343: Based on the real-time output feature vector and the dynamic feature vector of the desired target, construct a multi-target bias metric space, and calculate the Euclidean distance and cosine similarity of the output of each strategy execution instance and the dynamic desired target in different dimensions within this space; Step S344: Based on Euclidean distance and cosine similarity of the included angle, calculate the comprehensive deviation score of each policy execution instance in multiple dimensions through a preset weighted fusion rule; Step S345: Perform moving average and trend difference processing on the comprehensive deviation score according to the time window to remove instantaneous noise interference and generate a dynamic performance evaluation sequence of the strategy instance.

8. The method for providing scientific and technological information services based on cloud computing according to claim 1, characterized in that, Step S4, which involves encrypting and encapsulating the technology information recommendation strategy and its real-time performance evaluation parameters and binding them to a smart contract, includes the following steps: Acquire the science and technology information recommendation strategy and its real-time performance evaluation parameters, encapsulate them with the strategy generation timestamp and execution environment hash value to generate the original strategy data block; The encryption algorithms of multiple consensus nodes in the blockchain network are invoked to perform multi-signature and hash operations on the original policy data block, generating tamper-proof data fingerprints and node consensus proofs; Based on data fingerprints and node consensus proofs, a smart contract code segment containing policy content, fingerprints, proofs, and triggering conditions is constructed, and this smart contract code segment is bound to the original policy data block to generate a smart contract transaction to be confirmed. The smart contract transactions to be confirmed are broadcast to the blockchain network. After consensus is reached through the proof-of-work mechanism, they are packaged and linked to the blockchain to generate a trusted service strategy transaction block. Based on the distributed ledger characteristics of blockchain, trusted service strategy transaction blocks are synchronized to all service nodes. After each node verifies the validity of the block, it triggers the automatic execution logic of the smart contract to complete the service feedback to the requesting terminal.

9. The method for providing scientific and technological information services based on cloud computing according to claim 8, characterized in that, After synchronizing the trusted service strategy transaction block to all service nodes, the process also includes: After receiving the trusted service strategy transaction block, the service node parses the block header information and verifies the link relationship between its hash value and the previous block, as well as the validity of the multi-signature. After the block verification is successful, the node virtual machine loads and executes the smart contract code segment, first decrypting and extracting the technology information recommendation strategy from the encrypted original policy data block; Based on the strategy, the smart contract automatically generates a service response message that conforms to the interface specification of the requesting terminal, and pushes the service response message to the corresponding requesting terminal through the secure channel of the blockchain network. The requesting terminal confirms receipt of the service response within a preset time and generates a confirmation receipt containing a receipt timestamp, then broadcasts the confirmation receipt back to the blockchain network as a new transaction. After monitoring the confirmation receipt, the smart contract extracts the receiving timestamp and service policy identifier from the receipt to generate a service completion evidence tag. Based on the service completion evidence tag, it retrieves the interaction logs corresponding to the entire service process from the blockchain's temporary storage area, including on-chain records of policy generation, block confirmation, and policy push. The integrated service completes evidence labeling and interaction logs, and constructs a traceable event sequence for the entire lifecycle of this service by aligning timelines and associating events. Features are extracted from the traceable event sequence to obtain service time consumption, number of interaction rounds, and terminal response latency feature parameters. Combined with the technology information recommendation strategy and its real-time performance evaluation parameters, a multi-dimensional performance profile of the service is generated. The service multi-dimensional performance profile, technology information recommendation strategy, service response message and confirmation receipt are structured, recombined and compressed and encrypted to generate a complete service case data package, and its hash value is used as a unique index and stored in the on-chain storage area of ​​the historical technology information service case library.

10. A cloud computing-based science and technology information service system, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, for performing the cloud computing-based scientific and technological information service method as described in any one of claims 1-9.

Images