1005 results about "Large scale data" patented technology

The invention discloses an RAG enhanced Text-to-SQL (Structured Query Language) query method and system for a large-scale database environment. The method comprises the following steps: constructing a vector database; based on the original query of the user, generating query enhancement description aiming at Schema recall and query enhancement description aiming at SQL (Structured Query Language) by utilizing LLM (Logistics Language Model); carrying out Schema recall and historical question and answer pair recall operations in a double-way parallel manner by utilizing an RAG technology and relying on the constructed vector database; based on query enhancement description, recalled Schema and historical question and answer pairs, generating an SQL query statement by using LLM in combination with RAG, and realizing Text-to-SQL conversion; the generated SQL query statement is subjected to post-processing optimization through a database interface, the post-processing optimization comprises grammar verification, performance optimization and error correction, the SQL query statement is combined for execution and feedback, efficient, accurate and reliable natural language query is achieved, and continuous optimization can be conducted through user feedback.

The invention discloses a large and small model-based data anomaly detection method, system and device and a medium, belongs to the technical field of data processing and artificial intelligence, and aims to solve the technical problem of how to combine the rapid screening capability of a small model and the deep analysis capability of a large model to realize efficient and collaborative anomaly detection and cleaning. According to the technical scheme, an improved Isolation Forest small model is used for rapid screening, potential anomalies in large-scale data are rapidly detected through a dynamic threshold adjustment mechanism, adaptive feature weight distribution and multi-scale division optimization, an anomaly score is generated, and suspected abnormal data are preliminarily screened; secondly, performing fine detection on suspected abnormal data by utilizing a large model, and accurately identifying an abnormal type and generating a high-reliability abnormal result through deep semantic embedding analysis, abnormal feature clustering and domain rule verification; and finally detected abnormal data are classified and labeled, and the data characteristics are combined to process the abnormal data by adopting a deletion, correction or marking cleaning strategy.

The invention discloses a data center operation and maintenance fault prediction system and method based on deep learning. The system comprises a multi-source heterogeneous data acquisition module, a data preprocessing module, a deep learning prediction model module and the like. The method comprises the following steps: acquiring multi-dimensional operation data of a data center through full-quantity acquisition of multi-source data, and inputting a CNN-LSTM-Attention hybrid model to realize fault prediction after preprocessing and feature enhancement; fault grades are divided in combination with fault grading, early warning is pushed in multiple channels, a coping strategy is intelligently generated, the effect is verified in a closed loop mode, and finally the model is iteratively optimized. According to the scheme, the fault prediction precision and real-time performance are improved, the operation and maintenance response time is shortened, the service interruption risk caused by faults is reduced, and the method is suitable for efficient operation and maintenance of large-scale data centers.

The present invention relates to the technical field of blockchains. Disclosed are a smart-contract vulnerability detection method and system based on a hypergraph and multi-teacher distillation. Specifically, the smart-contract vulnerability detection method based on a hypergraph and multi-teacher distillation comprises: using a vulnerability detection open-source tool to perform vulnerability detection and analysis on source codes of a smart contract, and marking labels for the source codes of the smart contract; then, using a hypergraph to dynamically capture a deeper and richer relationship between the source codes, key functions and key variables in the smart contract, and on the basis of the hypergraph, training a plurality of vulnerability detection teacher models having excellent performance; and finally, using the plurality of vulnerability detection teacher models to perform knowledge distillation, so as to obtain a vulnerability detection student model, such that the vulnerability detection student model can also obtain knowledge without exposure to large-scale data, and thus directly applies the knowledge to vulnerability detection of new data. The smart-contract vulnerability detection method based on a hypergraph and multi-teacher distillation in the present invention can effectively detect various types of smart-contract vulnerabilities, and can effectively reduce the occupancy of storage space and improve the running speed.

The invention relates to a traffic anomaly detection and network security reinforcement method for a full data center, and the method comprises the steps: achieving the precise recognition and active blocking of an attack chain through the real-time collection of full network traffic, the construction of a multi-dimensional time sequence model based on historical attack chain features, and the combination of a dynamic judgment threshold value and spatial and temporal distribution feature analysis. The method specifically comprises the steps of analyzing traffic in real time to generate an attack chain feature sequence; performing clustering and correlation analysis based on the multi-dimensional time sequence model; dynamically adjusting the monitoring density to focus the high-risk area; and generating a threat blocking instruction and executing network isolation. Aiming at the problems of high attack chain omission ratio, resource contention conflict and insufficient dynamic defense capability caused by dependence on a static rule and a fixed threshold value in the traditional scheme, the method solves the problem of weak perception capability of the traditional technology on a complex attack chain, remarkably reduces the omission rate and the error blocking rate, guarantees the service continuity through dynamic resource scheduling, and improves the service performance of the system. The method is suitable for real-time security protection of a large-scale data center.

According to the OpenCSG large model intelligent data annotation system provided by the invention, by virtue of an automatic annotation technology driven by a large model, the requirement of manual annotation is greatly reduced, and the data annotation efficiency is remarkably improved. The system can generate a high-precision labeling result by using the intelligent labeling module and the automatic quality control module, and ensures the consistency and accuracy of the labeling result through an automatic auditing process. No matter sentiment analysis, entity recognition or complex text classification, the system can efficiently adapt to various types of labeling tasks, and different data processing requirements are met. Through modular design, the system supports flexible expansion, and can meet various data labeling requirements from small-scale test projects to large-scale commercial projects. According to the overall architecture, all the function modules are closely coordinated in the system, so that efficient operation and high-quality output are always kept in a large-scale data annotation project.

The invention discloses a distillation SAE and dynamic integration converter steelmaking carbon temperature soft measurement method, which comprises the following steps of: training a teacher model on a large-scale data set of a conventional production working condition, and compressing the model by adopting a knowledge distillation technology; then, a small amount of data from unconventional production working conditions is used for fine tuning to obtain a plurality of expert-type student stacking auto-encoder models SAE; in the prediction stage, a to-be-tested sample is mapped to respective feature space through a plurality of SAEs, and the posterior probability of the sample is calculated in each feature space based on a Gaussian distribution model; and finally, respectively inputting a to-be-measured sample into a plurality of corresponding regression devices to obtain a predicted value, and carrying out weighted fusion on the output of each regression device according to a posterior probability to realize dynamic integration selection of the soft measurement model. The method can effectively adapt to complex data distribution of multi-working-condition changes in converter steelmaking, the carbon temperature prediction precision under the unconventional working condition is improved, and the method has high robustness and engineering practicability.

The invention relates to the technical field of data processing, in particular to a computer feature extraction processing system and method fusing multi-source heterogeneous data, and the system comprises a structure correlation mapping module, a density mutation recognition module, a disturbance path analysis module, a path logic matching module and a risk logic output module. According to the method, by optimizing a data processing mode, important features in data can be accurately recognized, manual intervention and experience dependence are reduced, large-scale data analysis and processing efficiency is improved, the accuracy of a data model is enhanced, and by deeply analyzing the fluctuation amplitude and the change trend of the data, potential associated areas and abrupt change grids can be accurately positioned; key disturbance information is effectively identified, path logic matching can reveal a risk behavior mode in data evolution, potential early warning events are marked in time, the efficient processing capacity of the model for a complex data mode is ensured, and particularly, the problems of low efficiency and insufficient precision in a traditional scheme are solved in the face of large-scale data.

The invention provides a report generation and optimization method and system based on an offline data warehouse. The system comprises a data extraction layer, a data loading layer, a data cleaning layer, a data integration layer, a model data layer, a report mart layer, a report generation layer and a report optimization layer. The method comprises the following steps: S1, extracting data; s2, data cleaning; s3, integrating the data; s4, ensuring data quality; s5, storing model data; s6, a report market is constructed; and step S7, report generation. A layered architecture is adopted, independent operation of each module of the system is ensured, and expansion and maintenance are facilitated. Data extraction, cleaning and integration processes are simplified by introducing an automatic tool, and redundant calculation is reduced by optimizing SQL and solidifying logic. Bottom logic is optimized for large-scale data volume and complex query, and report generation performance is improved. And monitoring and management functions of the whole life cycle of the data flow are provided, and data quality and report accuracy are ensured.

The invention discloses an information management method based on a big data processing technology, and the method comprises the steps: S1, data collection and pre-processing optimization: S1.1, multi-source data cleaning and pre-processing: automatically detecting and correcting errors, missing and abnormal values in data through an integrated intelligent data cleaning technology, such as a machine learning algorithm (such as abnormal detection and missing value filling); the data quality is ensured, data standardization and format unification tools are adopted, heterogeneous data from different sources are converted into a unified format, and follow-up processing is facilitated. The problems of data privacy, quality, real-time performance, complexity and compliance in a traditional big data method are solved by enhancing data quality management, improving data security, optimizing storage and processing architecture and introducing modes such as real-time processing and intelligent analysis. Meanwhile, the expandability and the fault-tolerant capability of the system ensure the high efficiency and the stability of the system in a large-scale data environment, and a more intelligent, safer and more efficient information management solution is provided for enterprises.

A large language model generation method based on retrieval enhancement belongs to the field of computers and comprises the following steps: retrieving similar document abstracts according to input questions, converting the input questions and the document abstracts into points in a high-dimensional vector space, retrieving related documents and converting the related documents into a structured knowledge graph, generating sub-questions and extracting key information from the sub-questions, then, calling a large language model to carry out few-sample learning so as to generate associative cue words; generating a query command in combination with the associative cue word and the sub-question; obtaining sub-knowledge graph information by executing a query command, converting the sub-knowledge graph information into a text, and compressing and converting the text; and inputting the converted text and the question into a GPT-4 large language model to generate a natural language answer. According to the method, multi-level information and complex data in complex questions can be processed more accurately, the accuracy, efficiency and quality in large-scale data set and complex document processing are improved, accurate and coherent answers are provided, and the operation cost and resource consumption are reduced.

The invention discloses a macro server system, an application method, electronic equipment and a storage medium, and relates to the technical field of computers, physical devices (equipment components) of the macro server system are modularly designed to be combined into modules with different functions, and the physical devices support hot plugging, so that the physical devices can be conveniently and quickly combined into modules with different functions. Meanwhile, a plurality of physical devices are connected through a high-speed communication network to form a distributed module cluster, so that a large number of computing resources, management resources and network resources can be packaged into an independent server environment, and the limitation of a current server in the aspects of processor performance, bandwidth, physical structure, hardware quantity, firmware resources and the like is solved; according to the technical scheme, the macro server system capable of supporting large-scale data processing, high-concurrency access and complex operation is constructed, so that the technical problem of how to improve the performance and expandability of the server can be solved, and the technical effect of constructing the macro server system capable of supporting large-scale data processing, high-concurrency access and complex operation is achieved.

The invention provides a data processing method and system based on cloud computing, belongs to the technical field of data processing, and realizes high concurrent processing and automatic load balancing by dynamically scheduling computing power through elastic resource allocation and a distributed computing framework and combining a storage and computing separation framework. The system supports cross-node disaster recovery backup and multi-layer encryption, and data security is guaranteed; and an on-demand payment mode is adopted, so that the hardware input cost is reduced, the resource utilization rate is improved, and the large-scale data processing efficiency and the system expandability are remarkably improved.

The present application relates to the technical field of large-scale data processing, and particularly relates to a parallel computing method and system suitable for large-scale data processing. The system comprises a task management unit, a distributed load balancing module, an elastic expansion architecture, an intelligent communication optimization module, and a resource monitoring unit, wherein the task management unit divides large-scale data into a plurality of sub-tasks by means of a task decomposer, and distributes the sub-tasks to computing nodes by means of a task scheduler and a priority distributor; the distributed load balancing module achieves global load balancing by means of a load sensing unit, a dynamic adjustment unit and a balance optimization unit; the elastic expansion architecture dynamically adjusts system resources by means of a node manager, a resource pool controller and an expansion decision-making device; the intelligent communication optimization module optimizes inter-node communication by means of a communication path planning unit, a bandwidth distribution unit and a delay compensation unit; and the resource monitoring unit monitors the system performance in real time by means of a performance collector, a state analyzer and an anomaly detector.

The invention discloses an air-ground cooperative vehicle networking distributed digital twin service credible sharing method, which comprises the following steps of: firstly, according to an air-ground cooperative traffic data type, performing hierarchical packaging and storage on operation data of a vehicle networking distributed digital twin system; secondly, an authoritative random equilibrium consensus algorithm is used, and a distributed edge digital twinning system of the air-ground cooperative Internet of Vehicles carries out consensus on blocks generated by data and uploads the blocks to a block chain; and finally, requesting a twinning service from the remote digital twinning system at the edge digital twinning system, and verifying the service record through a directed acyclic graph block chain and an interstellar file system, thereby realizing the trusted sharing of the service among the edge digital twinning systems in the air-ground collaborative internet-of-vehicles distributed digital twinning system. According to the method, the authenticity and tamper-proof capability of the operation data of the air-ground cooperative Internet of Vehicles distributed digital twin system in cross-node sharing are ensured, the access efficiency of large-scale data is improved, and technical support is provided for cooperation of the Internet of Vehicles and the unmanned aerial vehicle.

The invention relates to the field of power system intelligent monitoring, and discloses a surface temperature sensor and a cable joint temperature monitoring system, and the system comprises the surface temperature sensor which is used for obtaining the temperature data of a cable joint, and transmitting the data to a data processing module; the data processing module is used for receiving the temperature data from the surface temperature sensor, storing the temperature data, formatting the received temperature data, removing invalid data and preliminarily screening abnormal values; and the temperature modeling module is used for calculating temperature distribution of the cable joint based on a heat conduction model, and establishing a temperature distribution model of the cable joint through a heat conduction equation in combination with the effective data output by the temperature data preprocessing module. According to the method, an optimization technical scheme combining Fourier transform and dimension reduction calculation is adopted, efficient processing and analysis of temperature data are achieved, the system response speed is increased, and the method is more efficient especially in a large-scale data processing environment.

The invention discloses a real model consistency review system and method based on a large language model, according to the scheme, through integration of multi-source heterogeneous data, standardized preprocessing of real scene cloud or images is completed, and through combination of deep learning recognition and multi-scale feature analysis, accurate correspondence and intelligent comparison of a real model and a BIM component are achieved; performing semantic reasoning and logic verification by using a large language model, automatically detecting differences such as dimensional deviation, positioning deviation, missing and redundant components and the like, and generating structured difference data; furthermore, difference information is visually presented through graphical annotation and natural language description, and batch screening and multi-scale interaction are supported. Based on a difference analysis result, a standardized review report is automatically compiled, and statistical analysis, change trend tracking and multi-format export functions are provided. According to the scheme, the consistency review efficiency and reliability in a large-scale data environment can be guaranteed, and the quality management level of building construction and operation and maintenance stages can be effectively improved.

The invention discloses an intelligent data storage method based on a cloud platform, relates to the technical field of data storage management, and is used for solving the problem that cold and hot data identification is lagged. According to the method, the access behavior graph is constructed, the concurrent access coupling weight is calculated, the data object with high access frequency and deep path hierarchy is identified, the structure adjustment instruction is dynamically generated, the access link complexity and index hierarchy redundancy are reduced, the response delay and the popularity prediction error before and after the structure adjustment are combined, and the access efficiency is improved. According to the method, the scheduling efficiency of the hotspot data is improved, the adaptability to the change of a storage access mode is enhanced, the dynamic adjustment capability and the structure response performance are good, and the method can be used for realizing the closed-loop linkage of the access behavior perception, the structure reconstruction and the model optimization. Therefore, the method is more suitable for a large-scale data intelligent management scene under a cloud platform.

The invention relates to the technical field of ocean information processing, in particular to a mesoscale vortex space-time characteristic analysis method and device based on a neural network, and the method comprises the following steps: obtaining the type, vortex center position, vortex intensity and temperature and sound velocity abnormal characteristics of each layer of mesoscale vortex according to the ocean current three-dimensional speed data of a target region; and the trained neural network predicts the vortex center position, vortex intensity and temperature and sound velocity abnormal characteristics of the mesoscale vortex at the future moment according to the type, vortex center position, vortex intensity and temperature and sound velocity abnormal characteristics of the mesoscale vortex at each layer at the historical moment. The problems that in the prior art, small-scale vortexes or rapidly-changing vortex characteristics are difficult to capture, large-scale data processing requirements are difficult to meet, and robustness is poor are solved.

According to the distributed clustering model training device based on the insurance big data and the application thereof, the distributed clustering model training device based on the insurance big data and the application thereof are combined with distributed computing, cloud computing optimization and a big language model auxiliary strategy, and efficient data processing, low-cost computing resource scheduling and intelligent modeling are achieved. The device adopts a security mechanism to ensure data compliance, introduces man-machine collaborative optimization, improves model suitability and prediction precision, and meets the requirements of the insurance industry for large-scale data analysis.

The invention discloses an accurate analysis method for Internet user behaviors based on artificial intelligence, and the method belongs to the technical field of Internet, and comprises the steps: collecting user behaviors and basic attribute information through a multi-source data collection technology, and integrating data through a data fusion algorithm after data cleaning and preprocessing; designing and extracting a plurality of behavior features, screening key features by using a feature selection algorithm, and performing feature transformation processing at the same time; constructing a deep learning model fusing a recurrent neural network (such as LSTM and GRU), a convolutional neural network and the like, training the model by using a large-scale data set, and optimizing parameters; establishing a real-time analysis system, inputting new data into the model in real time, providing personalized service and recommendation according to an analysis result, and collecting feedback data to update the model online; and finally, evaluating the model by adopting multiple evaluation indexes, verifying the generalization ability through methods such as cross validation and the like, and further optimizing and adjusting the model.

The invention relates to the technical field of personalized recommendation, in particular to a nonlinear enhanced decoupling type contrast hypergraph learning method for next POI recommendation. The method comprises the following steps: S1, constructing a multi-view decoupling hypergraph; s2, optimizing the multi-view decoupling hypergraph constructed in the step A to obtain a nonlinear hypergraph convolutional network; s3, based on self-adaptive fusion of user representation, learning and fusing the user preferences under the multiple views obtained in the step S2; and S4, realizing comparative learning and self-supervised learning based on an Encoder module and a GRACE module. According to the nonlinear enhanced decoupling type contrast hypergraph learning method for the next POI recommendation, by introducing ReLU and residual connection, the linear limitation of a traditional hypergraph layer is broken through, the model is endowed with higher expression ability, a KNN adjacency matrix sparse strategy is adopted, the calculation efficiency is remarkably improved on the premise that the precision is guaranteed, and the method is suitable for popularization and application. And meanwhile, batch processing InfoNCE and GRACE modules are creatively applied, so that robust cross-view collaborative learning on a large-scale data set is realized.

The invention relates to the technical field of automatic driving, and discloses an unmanned vehicle navigation method based on multi-modal fusion, which comprises the following steps: S1, acquiring and preprocessing sensor data, acquiring multi-modal sensor data, and performing data alignment, filtering and normalization processing; s2, constructing an optimization framework of the fusion weight of the sensor based on a variational inference method, defining an optimization objective function by using the preprocessed data as input, and solving an optimal solution of the fusion weight through the variational inference method; and S3, on the basis of the variational reasoning optimization model, optimizing a sensor weight updating process by utilizing information geometric measurement. A cross-modal learning and reasoning mechanism is introduced, weight updating is accelerated by using an information geometric optimization method, so that the unmanned vehicle can quickly adapt to sudden environmental changes, and compared with a traditional neural network model needing a large amount of training data, the method does not need large-scale data set pre-training, can adaptively adjust decisions, and improves generalization ability.

The invention relates to the technical field of data processing, and discloses a function expression heterogeneous data source processing method and device.The method comprises the steps that a first query statement and a data processing expression in a data processing text are obtained, and the first query statement is set according to the same self-defined grammar rule; converting the first query statement into a corresponding abstract syntax tree according to a preset analysis rule, traversing nodes of the abstract syntax tree to generate a second query statement, querying the corresponding databases based on the second query statement, obtaining target data of a plurality of heterogeneous databases, and performing integration calculation on the target data by using an expression. According to the method, the first query statement only has one rule, after the first query statement is converted into the second query statement, the second query statement can be matched with the respective database, and research personnel only need to learn the grammar rule of the first query statement to simplify the process of manual data calculation; and the data processing efficiency during large-scale data integration based on the heterogeneous data source is improved.

The invention relates to the technical field of computer programming, and discloses a Go language structural body automatic conversion method and system. The method comprises the following steps: firstly, receiving definitions of a source structure body and a target structure body as input, automatically matching fields of the two structure bodies through a field mapping rule engine according to a preset rule, and generating a field mapping relationship; then, based on the field mapping relation, a code generator is used for automatically generating a conversion code containing field assignment and type conversion logic; finally, the generated conversion codes are compiled into an executable conversion function, and the function can convert data in the source structure instance according to the field mapping relation during program running and fill the data into the target structure instance. According to the method, automation of structural body conversion is achieved, the problem that errors are prone to occurring when conversion codes are written manually is solved, and the maintainability of the codes is improved; the mode of reflection during operation is replaced by code generation, so that the conversion efficiency is remarkably improved; meanwhile, the method supports batch conversion and concurrent processing, and can meet the performance requirements of large-scale data processing.

The invention discloses an intelligent substation configuration file checking method and system based on artificial intelligence, and relates to the field of artificial intelligence, and the method comprises the steps: responding to an input signal of a to-be-checked configuration file, extracting n character strings of the to-be-checked configuration file, carrying out the semantic analysis of each character string, and obtaining a semantic analysis result, n is an integer greater than 1; key features of the analysis result are extracted from the semantic analysis result, feature learning is conducted on each key feature through a preset feature learning algorithm, and high-level feature representation of each key feature is obtained; combining all high-level feature representations of the analysis result to obtain a feature set; and inputting the feature set into an artificial intelligence model to obtain a proofreading result of the to-be-checked configuration file and a preset reference configuration file. According to the method and the device, the problem of huge resource consumption caused by a traditional checking mode can be solved in the face of large-scale data or complex comparison requirements.

The invention discloses a data privacy protection method for an industrial internet platform, and relates to the technical field of data security and privacy protection. According to the method, the sensitive information is accurately identified and deeply analyzed through the sensitive information feature library and the hierarchical matching algorithm, the problem of insufficient accuracy and flexibility during large-scale data processing is solved, the accuracy and reliability of data desensitization are remarkably improved, personal privacy is effectively protected, data availability is maximized, and the method is suitable for large-scale data processing. A hierarchical processing algorithm and a self-adaptive desensitization rule base are utilized, desensitization rules are dynamically adjusted according to dynamic access requirements and sensitivity levels of data, data security and availability are balanced, accurate protection under different scenes is ensured, system adaptability and flexibility are improved, and the data access process is monitored in real time, anomaly detection and rule verification are performed, so that the data access efficiency is improved. And the desensitization rule is dynamically adjusted, so that the security and reliability of the system are enhanced, the user credibility is improved, and the transparency and credibility of the data processing process are ensured.

The invention discloses a wind power plant intelligent video monitoring and abnormal behavior recognition system, method and device and a medium, and belongs to the technical field of wind power generation, and the system comprises a multi-mode collection module which is used for collecting various data of a wind power plant; the edge calculation module is used for receiving and primarily processing data of the wind power plant; the communication network module is used for transmitting the processed data of the wind power plant to the central server cluster module; the central server cluster module is used for receiving the data transmitted by the communication network module and executing deep learning model training and large-scale data analysis; and the unmanned aerial vehicle subsystem is used for executing large-range inspection and close-range inspection and transmitting the acquired data to the central server cluster module through the communication network module. The problems that a traditional wind power plant monitoring system is difficult in data integration, weak in analysis capacity and poor in real-time performance are solved, the operation efficiency, safety and reliability of a wind power plant are remarkably improved, and technical support is provided for sustainable development of the wind energy industry.

The invention relates to the technical field of big data analysis and disaster management, and discloses a multi-dimensional spatio-temporal data analysis and visualization method, system and device. The method comprises the steps of obtaining multi-source spatio-temporal data and performing standardization processing to obtain processed spatio-temporal data; performing real-time analysis and trend prediction according to the processed spatio-temporal data based on a deep learning algorithm to obtain an analysis prediction result; performing three-dimensional modeling through a GIS platform according to the processed spatio-temporal data and an analysis prediction result to obtain a three-dimensional dynamic simulation model and performing visual display; and generating a decision strategy according to an analysis prediction result, and optimizing the decision strategy based on a reinforcement learning mechanism. According to the method, multi-source spatio-temporal data are efficiently integrated, real-time processing of large-scale data is achieved, and the data response speed is remarkably increased. Prediction is carried out based on multi-source data fusion and a deep learning algorithm, and the automation and accuracy of analysis are improved. Real-time dynamic display helps a user to understand a data trend more visually, and decision scientificity and timeliness are enhanced.

The invention discloses a semi-supervised learning method based on a self-adaptive threshold value, belongs to the field of radio communication, and aims to extract general semantic features of signals through comparative learning so as to improve generalization of downstream modulation identification tasks. A self-adaptive threshold mechanism is introduced to generate a high-quality pseudo tag, and confirmation deviation is reduced through additional category information, so that the robustness of the model is improved; besides, a hierarchical encoder capable of learning Fourier filtering is designed to capture multi-scale semantic features of large-scale long-sequence signals, improve the calculation efficiency of a model and research and verify an algorithm based on a plurality of large-scale data sets, so that the accuracy of automatic modulation recognition is remarkably improved, and particularly, under the condition of less label data, the accuracy of automatic modulation recognition is greatly improved. Compared with a mainstream method, the method has higher generalization ability and robustness, and can effectively cope with signal diversity and noise interference in a real communication environment.