473 results about "Sorting algorithm" patented technology

The invention relates to a task scheduling optimization method and device based on reinforcement learning, equipment and a medium. The method comprises the steps that firstly, system resource state data are collected in real time, dynamic environment characteristics are determined through preprocessing and time sequence analysis, task characteristic data are analyzed at the same time, and a task priority sequence and a resource demand vector are generated through a priority ranking algorithm and a resource evaluation model; and then a state space and an action space are constructed by adopting a reinforcement learning algorithm, an optimal task allocation scheme is generated through strategy iteration and reward function optimization, and if the scheme meets a resource balance threshold, scheduling is executed, and performance indexes are collected. And finally, fusing real-time indexes with historical data, and updating parameters of the reinforcement learning model through experience playback and gradient descent to form a closed-loop optimized improved scheduling strategy. By adopting the method, the accurate mapping of the resource state and the task requirement can be realized, and the problem of insufficient adaptability of the traditional static scheduling to a complex scene is solved.

The invention discloses a transmission network fault positioning method, system and device and a storage medium, and relates to the field of communication, and the method comprises the steps: firstly, collecting network data to construct a multi-granularity network topology model, then obtaining original alarm data, analyzing the topology association relationship between alarms, and obtaining a multi-granularity network topology model; and mining horizontal peer association and vertical upstream and downstream association among the alarm objects, and identifying multi-alarm data of each alarm object. Thirdly, constructing an associated alarm directed graph based on the multi-alarm data, and performing associated alarm compression on the associated alarm directed graph to obtain a compressed alarm set; and finally, constructing an alarm propagation graph based on the compressed alarm set and the multi-granularity network topology model, querying the alarm propagation graph through a reverse hierarchical sorting algorithm, and determining a root alarm. According to the method, through innovative methods such as topological modeling, association mining, graph compression and reverse sequencing, alarm data and a topological structure are closely combined, an alarm propagation rule is analyzed, and fault root causes are diagnosed.

The invention discloses a milk industry knowledge question-answering method and device based on a large model and RAG and a medium, and the method comprises the steps: collecting multi-source heterogeneous data of the milk industry, carrying out the term protection word segmentation processing of the multi-source heterogeneous data, and generating a labeled field corpus; based on a pre-training language model base, through a domain corpus injection and adversarial learning mechanism in a domain corpus, generating an enhanced language model adapted to the dairy industry terminology; receiving a milk industry problem of a user, executing semantic vector retrieval and term extension retrieval in parallel, and screening a multi-modal retrieval result through a dynamic sorting algorithm; and splicing the multi-modal retrieval result and the milk industry question into an enhanced prompt, inputting the enhanced prompt into an enhanced language model, generating a final answer corresponding to the milk industry question, and associating the final answer with a knowledge source.

The invention relates to the technical field of equipment maintenance scheduling, in particular to an equipment fault repair management system, which comprises a dependency identification and modeling module, a fault state analysis module, a task priority evaluation module, a work order scheduling generation module and a resource scheduling execution module. According to the method, a multi-level dependency topology between devices is constructed through a directed graph traversal algorithm, a trigger relation between physical connection and an operation process is quantitatively analyzed, a classification marking model is established in combination with dynamic parameters such as a performance degradation rate, a fusion influence range and a dependency factor are calculated through scalar superposition, and a priority scoring matrix is dynamically generated. A task queue structure is optimized through a sorting algorithm, a data-driven maintenance decision mechanism is formed, the fault positioning precision is improved, response delay caused by manual intervention is reduced, key node equipment maintenance lag is avoided, the resource configuration efficiency is optimized, the collaboration of fault processing and a production system is strengthened, and formulation of a preventive maintenance strategy is supported.

The invention relates to a high-fidelity cloud rendering cluster scheduling method and system, and the method comprises the steps: receiving a rendering task, generating a multi-level priority queue according to the task complexity, timeliness and resource demand classification, and automatically optimizing a scheduling strategy; monitoring the resource state of the heterogeneous computing node in real time; allocating tasks by using preemptive and round-robin scheduling strategies, and adjusting and coping with resource fluctuation in combination with a dynamic code rate; the tasks are decomposed by adopting a spatial blocking and time framing strategy, and an execution sequence is controlled according to a topological sorting algorithm; abnormal nodes are identified through heartbeat detection, and affected subtasks are migrated through incremental task updating. According to the method, efficient resource allocation, scheduling algorithm and idle key frame scheme can be realized, the GPU directly outputs the video stream, the rendering speed is remarkably improved, resource waste and operation cost are reduced through a dynamic resource allocation mechanism, a fault-tolerant mechanism is provided, the task is ensured to be normally completed under the condition of node fault, and the task efficiency is improved. And distributed rendering and synthesis of large-scale high-resolution images are supported.

The invention provides a transfer learning-based power transmission line fault positioning method and system. Real-time monitoring data of a target power transmission line are collected, meteorological information and satellite remote sensing images are combined, ambient environment changes are analyzed, anomaly detection is performed, an abnormal data set is obtained, and a comprehensive fault feature space is constructed; an adversarial domain adaptation algorithm is adopted to optimize a comprehensive fault feature space, and a long-short-term memory network, an attention mechanism and a graph convolutional network are used for data processing to generate an enhanced fault feature representation; constructing an integrated learning framework based on Bayesian optimization, predicting the target power transmission line by using the plurality of preliminary basic models, integrating all prediction results through an adaptive weighting strategy, and generating initial fault probability distribution; determining fault positioning information by using a geographic information system technology and a topological sorting algorithm; the invention provides an efficient and accurate power transmission line fault positioning method, and provides powerful support for safe operation of a power system.

The invention provides an industrial data online migration method, medium and system of a productivity middle platform, and belongs to the technical field of electrical digital data processing.The method comprises the steps that firstly, historical access records of an industrial data source are collected, and a deep neural network model comprising a multi-head attention layer, a time sequence feature extraction layer and a residual connection layer is constructed; and the prediction of the data change rate is realized. And based on a prediction result, dynamically calculating a data change rate demarcation point by utilizing an optimization equation set, and dividing a data table into different change frequency categories. And then constructing a data dependency relationship matrix, and determining a migration sequence by adopting an improved topological sorting algorithm. Batch migration, incremental migration and real-time synchronization mechanisms are adopted for different types of data, and model updating is triggered through real-time monitoring, so that dynamic optimization of a migration strategy is ensured. Finally, data consistency verification is carried out, migration accuracy is guaranteed, and the technical problem that the online migration efficiency of industrial data is low due to real-time dynamic change of the data change rate is solved.

The invention relates to the technical field of association retrieval, and discloses an internal and external network knowledge association retrieval method based on an RAG technology. The method comprises the following steps: S1, based on query content input by a user, retrieving matched entries from an enterprise internal knowledge base, and reducing a retrieval range through metadata filtering; and S2, converting internal retrieval results into vector representation through an embedded model, and sorting the results according to correlation on the basis of cosine similarity. By combining an enterprise internal knowledge base and external network search, the limitation that a traditional method only depends on a single data source is broken through, the coverage range of knowledge is remarkably expanded, professional and authoritative domain knowledge is provided for the enterprise internal knowledge base, and latest industry dynamic and open information is supplemented by external network search, so that the enterprise internal knowledge base and the external network search are combined, and the enterprise internal knowledge base and the external network search are combined. Through dynamic weight adjustment and a unified sorting algorithm, the most relevant data source can be preferentially selected according to user requirements, so that balance is achieved between content depth and breadth.

The invention provides an automatic incremental compiling method and system for Java development, and the method comprises the steps: carrying out the hash value verification of an obtained Java source code, and obtaining the change event information in the Java source code; extracting static dependency information and a dynamic calling relation from the change event information, and generating a relation dependency graph according to the static dependency information and the dynamic calling relation; based on the relation dependency graph, generating a compiling task queue by utilizing a topological sorting algorithm; compiling the change event information in the Java source code according to the compiling task queue to obtain a machine code corresponding to the change event information; according to the method, code change information is accurately positioned through hash value verification, so that the initial range of incremental compiling is favorably reduced; according to the method, the static dependency information and the dynamic calling relation are extracted from the change event information, the relation dependency graph is generated, the dependency relation between the code units is completely represented, the problem that the dynamic calling relation is omitted is solved, and compiling correctness and efficiency are improved.

The invention provides a 3D + 2D high-dimensional feature fusion-based material tracking multi-target identification method and system. The method comprises the steps of performing target identification on a to-be-tracked material by using a YOLO algorithm; performing multi-target real-time tracking on the material based on a Deep SORT algorithm; generating three-dimensional point cloud data of a target by adopting a multi-view data calibration and alignment algorithm, and carrying out noise reduction, sampling and data enhancement preprocessing operation; performing three-dimensional reconstruction on the preprocessed three-dimensional point cloud data by using a Gaussian Splitting method, and introducing time sequence information at the same time; segmenting a dynamic three-dimensional target and a static background through a Mask R-CNN technology; aligning the position of the target in the two-dimensional image with the coordinates in the three-dimensional space by using a feature fusion method; a PNP algorithm is adopted to estimate attitude and depth information of a target in a three-dimensional space, accurate positioning and state restoration of materials are realized, and efficient and reliable technical support is provided for material management and tracking in an industrial environment.

The invention relates to the technical field of photovoltaic systems, and particularly discloses a fault identification method and system for a photovoltaic system, and the method comprises the steps: collecting data in real time through environment, electrical parameters and an equipment state monitoring sensor, carrying out the cleaning and standardization, extracting time domain, frequency domain and time frequency features, and screening a feature subset through a correlation analysis and feature importance sorting algorithm; detecting and classifying faults by using a hybrid model composed of an isolated forest algorithm and a random forest classifier; positioning a fault subsystem and analyzing a root cause by means of a hierarchical diagnosis strategy, a graph neural network and a Bayesian reasoning algorithm; and periodically updating the model based on the new data. The method can quickly and accurately identify and position faults, adapts to a complex environment, reduces the operation and maintenance cost, and improves the operation reliability and operation and maintenance efficiency of a photovoltaic system.

The invention provides a multi-agent system initialization method and device for a question and answer large model, and relates to the technical field of question and answer large language model multi-agent systems. The method comprises the following steps: disassembling a user question and answer query task into a plurality of sub-tasks by adopting a pre-constructed question and answer planning agent, generating an agent corresponding to each sub-task, carrying out standardization processing through a formatting agent to obtain a standardized agent corresponding to each sub-task, and carrying out evaluation through observing the agents to obtain feedback information. Generating a candidate agent set based on multiple iterations; a pre-training text encoder is adopted for encoding, and semantic representation and task representation of each agent are generated; task correlation and agent diversity are calculated, and a multi-objective optimization problem is constructed; and obtaining a Pareto optimal frontier team set by adopting a non-dominated sorting algorithm, and selecting an optimal agent team set by selecting agents. According to the invention, the cooperation efficiency of the multi-agent system can be improved.

The invention discloses a reading sequence detection method and system combining layout analysis and a language model. The method comprises the following steps: collecting a historical handwritten document image and preprocessing the historical handwritten document image; performing layout analysis on the document image, and identifying and classifying interested text areas in the document image; performing semantic similarity evaluation on different types of layout entities in combination with a language model; and detecting a reading sequence by using a sorting algorithm through document layout and semantic information. The method has the advantages that the reading sequence detection method for the historical handwritten document image is provided through combined recognition of the text and the layout, and conversion of the document from the image to the text content conforming to the reading sequence is achieved.

The invention provides a household appliance knowledge question-answering method and system based on retrieval enhancement generation. The method comprises the following steps: acquiring household appliance field multi-modal data from a multi-format document library; extracting text information, table information and chart information in the multi-modal data; performing domain term injection processing on the extracted information, and constructing a packet domain enhancement index; receiving a natural language question input by a user; the natural language problem is analyzed through a query optimizer, and semantic retrieval and keyword retrieval are executed in parallel; carrying out fusion processing on the semantic retrieval result and the keyword retrieval result; selecting matched document fragments by adopting a relevancy sorting algorithm; inputting the matched document fragments into a large language model to generate candidate answers; verifying the compliance and traceability of the candidate answers through a credibility evaluation module; outputting a final answer with a reference source; and storing the high-frequency questions and the final answers into a cache library to solve the problems that the answer accuracy of a knowledge question-answering system is reduced and the response efficiency is limited.

The invention relates to the technical field of railway fault traceability, in particular to a railway fault traceability method and system based on a weighted path sorting algorithm, and the method comprises the steps: constructing a traceability knowledge graph according to historical fault data, equipment operation data and maintenance records of a railway system; and then a weighted path sorting algorithm is carried out on the traceability knowledge graph by combining a Bayesian network model, a random walk algorithm and an attention mechanism model, so that efficient and accurate railway safety fault traceability is realized, the safety and reliability of railway transportation are remarkably improved, and the method has a wide application prospect and a remarkable practical value.

The invention provides a user complaint processing method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining user complaint information; based on the first cue word, guiding the knowledge analysis large language model to determine key information, guiding the knowledge analysis large language model to retrieve the key information in the knowledge graph based on a path sorting algorithm, and generating a plurality of candidate solutions based on a retrieval result; performing sentiment analysis on the user complaint information based on a sentiment analysis large language model, and determining the sentiment level of the user complaint information; and based on a second cue word, guiding a knowledge analysis large language model, and based on the emotion level, determining a final solution from the plurality of candidate solutions. According to the method, an automatic complaint processing process from information extraction, knowledge retrieval to scheme generation and emotion adaptation is constructed through division and cooperation of the knowledge analysis large language model and the emotion analysis large language model, a judgment process needing manual intervention is avoided, and complaint processing efficiency is improved.

The invention discloses a short message communication method and system for FTU (feeder terminal unit) equipment of a power distribution network. The method comprises the following steps: acquiring a power state data stream, an equipment fault data stream and an equipment health state data stream in the FTU equipment of the power distribution network; each data stream is preprocessed; calculating the priority of each data stream by using a dynamic priority sorting algorithm based on each preprocessed data stream, and classifying the data streams into a high-priority data stream and a low-priority data stream; monitoring the real-time quality data of the communication link based on the Beidou No.3 satellite system, and predicting the health state of the satellite link by using an extended Kalman filtering algorithm to obtain a link health state prediction result; the real-time quality data comprises link delay, bandwidth and packet loss rate; and calculating a quality score of each link by using a nonlinear programming algorithm based on a link health state prediction result, and selecting the link according to the score to carry out data stream transmission. The reliability and the transmission efficiency of the system can be improved.

A large model agent generation content quality evaluation method based on double baselines comprises the steps that data generated by human beings is acquired, preprocessing and semantic segmentation are carried out, and data samples after semantic segmentation are formed; performing theme extraction on the data samples after data segmentation to form a theme list; the content of the data forming the theme list is generated through an intelligent agent workflow and the content of the data forming the theme list is directly generated through a large model, and an evaluated sample and a contrast sample are obtained; comparing and scoring the data sample after semantic segmentation, the evaluated sample and the contrast sample; selecting a sorting algorithm to carry out score conversion on the graded and compared samples; and performing statistical verification and result analysis. According to the method, a comprehensive, reliable and efficient solution is provided for quality evaluation of the generation content of the large model agent. The method not only can accurately evaluate the content quality, but also provides powerful support for continuous optimization of an intelligent agent system.

The invention discloses a large model-based government affair knowledge retrieval method, apparatus and device, and a medium, and relates to the technical field of natural language processing, and the method comprises the steps of performing information extraction on a to-be-processed government affair work order through a preset government affair large model to obtain target structured information; based on a word processing fusion algorithm constructed by a word frequency-inverse document frequency algorithm and a graph sorting algorithm, performing word processing on the target structured information to obtain work order keywords; converting the work order keyword into a target semantic vector; performing retrieval operation on a preset government affair knowledge base by utilizing a preset retrieval condition and the target semantic vector to obtain a target government affair knowledge document; the preset retrieval condition is a retrieval condition constructed based on time, space and the work order emergency degree. Therefore, the information accuracy and the processing efficiency can be improved through word processing; and in combination with retrieval conditions constructed based on time, space and work order emergency degrees, accurate retrieval can be provided, and the processing effect on various complex government affair work orders is improved.

The invention relates to an automatic patent document generation method and system based on retrieval enhancement, and the method comprises the steps: carrying out the feature extraction of text content inputted by a user, and obtaining keywords and key sentences of the text content; based on the keywords, performing text retrieval in a pre-constructed text database, and sorting text retrieval results; based on the key sentences, vector retrieval is carried out in a pre-constructed vector database, and vector retrieval results are sorted; reordering the text retrieval result and the vector retrieval result through a reciprocal ordering algorithm; determining an auxiliary enhanced document in the reordering result; inputting the text content input by the user and the auxiliary enhanced document into the large language model, and outputting a patent document conforming to the standard template through the large language model; wherein patent documents in a text format are stored in the text database; vectorized patent documents are stored in the vector database; the large language model is trained based on a text database and a vector database.

The invention relates to the technical field of embedded system construction, in particular to a software dependency relationship automatic integration method and system based on meta packaging, and the method comprises the steps: S1, creating a meta packaging framework in an embedded construction system, and defining and constructing a stage hook through a Makefile; s2, automatically deploying the pre-compiled software package and the shared dependency library to a target directory of a construction system through a resource injection script; s3, analyzing a metadata file of the pre-compiled software package, extracting dependency relationship information, and generating a global dependency relationship list based on a directed acyclic graph topological sorting algorithm; and S4, inversely writing the global dependency relationship list into the dependency declaration configuration of the meta-packaging by utilizing the dependency aggregation script. By implementing the scheme, the problems of dependency sequence disorder and metadata auditing failure during embedded system integration are solved, the reliability of system integration is improved, the system construction efficiency is optimized, the safety of system construction is enhanced, and the maintainability and expansibility of system construction are improved.

The invention relates to an optimization method for sorting an openGauss database. The method comprises the following steps of: acquiring various different types of data such as int, float, numeric, varlar and the like from a data source, carrying out Normalization processing on the acquired data, and then putting the data subjected to Normalization processing into a sorting space; sorting and comparing the data in the sorting space by using a memcmp function of a standard C library, and exchanging Normalize data participating in comparison according to a comparison result; and orderly data are directly obtained from the sorting space in sequence. According to the method, the data is directly sorted instead of being indexed through rowid, so that the data access performance and sorting efficiency of the openGauss database can be remarkably improved. According to the method, the data are all subjected to Normalization, and the data length is a fixed value, so that a sorting algorithm with lower time complexity can be used.

The invention discloses a rapid battery consistency sorting method based on an electrochemical impedance spectrum. The method comprises the following steps: S1, acquiring the electrochemical impedance spectrum of all batteries to be sorted; s2, inputting the electrochemical impedance spectrums of all the to-be-sorted batteries into a pre-trained principal component analysis conversion matrix, so that the high-dimensional impedance characteristics are reduced into one-dimensional impedance characteristics; and S3, inputting the one-dimensional impedance characteristics of all the to-be-sorted batteries into a greedy sliding window sorting algorithm to obtain a consistency sorting result of the to-be-sorted batteries. According to the method, the sorting characteristic measurement time is short, the sorting efficiency is high, the capacity utilization efficiency of the sorted battery module is higher, the service life is longer, and the method has important significance for improving the efficiency and effect of battery consistency sorting.

The invention discloses a fast charging station planning method based on urban dynamic traffic demand prediction and correction, and relates to the technical field of electric vehicle fast charging station planning, and the method comprises the steps: S1, constructing a traffic topology connection matrix according to urban road network topology, and recognizing traffic hub nodes and long-distance nodes as candidate nodes; s2, proposing three dynamic factors including social demands, user public praise and demand elasticity, and correcting the electric vehicle traffic demand prediction value of each path unit; s3, constructing a multi-target fast charging station capacity optimization model taking the minimum planning total cost and prediction error as targets, wherein the multi-target fast charging station capacity optimization model comprises fast charging station self constraints and power grid adaptation constraints; s4, solving the model by adopting a non-dominated genetic sorting algorithm to obtain a Pareto solution set, and selecting a planning result according to the demand of a decision maker; according to the method, the problems of blind site selection and static demand prediction in traditional planning are solved, economy and power grid safety are considered, planning scientificity and decision-making flexibility are improved, and supply and demand of charging resources can be effectively balanced.

The invention provides a user-defined script dynamic editing and executing method and system of a Unity engine. The method comprises the following steps that codes are input and preprocessed; executing text cleaning, word segmentation and annotation preprocessing according to the line feed as a code line set; grammatical analysis and semantic understanding; constructing an abstract syntax tree, identifying a syntax structure, and combining context analysis semantics to generate a symbol table; processing errors in real time; positioning an error line through a multi-stage matching strategy, highlighting an error position through a visual interface, and generating a repair suggestion in combination with a line number mapping table, error propagation analysis and a multi-evidence fusion mechanism; executing dynamic codes; explaining and executing the corrected script, and feeding back an execution result; through interaction with a Unity engine, operation control of a script on a game object and a scene is realized; through three core technologies of semantic weight dynamic matching, an anti-interference navigation state machine and a hybrid sorting algorithm, the long-standing problems of low matching precision, poor interaction experience and the like in the field of code completion are solved.

The invention discloses a cache tail delay optimization method based on NVMe solid state disk characteristics, and aims to solve the problem of cache long tail delay caused by a high fan-out request in a cache system. The method comprises a differential request processing strategy: based on weighted polling arbitration mechanism characteristics of the NVMe solid state disk, determining the number of sub-requests contained in different priorities according to fan-out sizes of requests, and reducing processing time differences among the sub-requests while giving higher priorities to high fan-out requests; according to the sub-request sorting algorithm, the influence of the sub-requests on the tail delay is evaluated according to the characteristics of the sub-requests, and sorting is performed according to the sub-requests, so that the priorities of the sub-requests are determined; according to the priority queue load balancing strategy, loads between NVMe priority queues are balanced, and queue congestion is avoided; and the feedback adjustment mechanism is used for monitoring and analyzing queue load and request delay to perform dynamic adjustment. Compared with an existing cache based on a solid state disk, the method has the advantage that the tail delay of the cache is optimized by fully utilizing the characteristics of the NVMe solid state disk.

The invention discloses a regional water environment bearing capacity change evaluation and prediction method and system, and the method comprises the steps: obtaining water resource data, water environment data and social economic data in a target region, and constructing a water environment bearing capacity index system; the index system comprises a plurality of subsystems; evaluating the subsystems through a projection pursuit model based on a sparrow algorithm to obtain subsystem evaluation results; obtaining a comprehensive evaluation result of the water environment bearing capacity through an entropy weight approximation ideal solution sorting algorithm according to the subsystem evaluation result; constructing a Markov transition probability matrix according to the comprehensive evaluation result; and predicting the change probability of the water environment bearing capacity according to the Markov transition probability matrix. The method can improve the evaluation and prediction accuracy of the water environment bearing capacity change, and can be widely applied to the technical field of water environment evaluation and prediction.

The invention relates to the technical field of medical care informatics, and discloses a medical questionnaire open question answer coding method, system and device and a medium, and the method comprises the steps: converting an answer document of a medical questionnaire open question into an answer document vector through a document embedding model; an aggregation hierarchical clustering algorithm is utilized to quickly and accurately identify answer document clusters with hierarchical relationships and different answer document themes in answer documents, so that complex information is structured, subsequent analysis is facilitated, and a richer view angle is provided for data analysis; in combination with a graph-based sorting algorithm for texts and an abstract text abstract model, encoding answer documents contained in answer document clusters of different answer document topics into answer document abstracts, and forming topic-abstract pairs with hierarchical relationships by the answer document topics and the answer document abstracts of the answer document clusters; manual intervention is reduced, human errors are avoided, the consistency and reliability of medical data processing are improved, and the medical data processing time is shortened.

The invention discloses a production control method and system based on multi-source data fusion and an intelligent algorithm, and relates to the technical field of automatic control. Multi-dimensional production state information is collected through a sensor, multi-source heterogeneous data is integrated into a data set in a unified format through a data fusion technology, and the multi-dimensional production state information is obtained; a long short-term memory network is used for trend analysis and prediction, a multi-objective optimization algorithm is used for calculating an optimal control parameter, multi-loop cooperative adjustment is realized, an adjustment instruction is issued through a distributed control protocol, a priority ranking algorithm is introduced to ensure system stability, and a game theory negotiation algorithm is used for realizing dynamic resource allocation among production units. The production scheduling plan is optimized through the genetic algorithm, continuous monitoring and optimization can be achieved, control parameters can be adjusted in a self-adaptive mode, the production efficiency and the resource utilization rate are effectively improved, and intelligent automatic control under the complex production environment is achieved.

The invention discloses a multi-dimensional natural language reasoning method based on a safety emergency knowledge graph. The method comprises the following steps: S1, collecting multi-source heterogeneous data, and carrying out standardized cleaning, entity recognition and relationship extraction; s2, constructing an emergency knowledge graph, performing storage and indexing by adopting a graph database, and designing an incremental updating mechanism; s3, intention classification and keyword recognition are conducted on natural language questions input by a user through the deep learning model, and the questions are mapped into initial entity nodes in the graph; s4, performing multi-hop reasoning around an initial entity in the knowledge graph, screening out a most relevant and logically coherent Top-K path in combination with reinforcement learning and a graph sorting algorithm, and realizing deep correlation analysis and answer candidate generation of the question; s5, generating an answer in a structured and natural language mixed form based on a reasoning result, and providing a visual reasoning path for display; a user feedback mechanism is introduced, and continuous optimization of the system and iterative updating of the model are supported.