148 results about "Sequence pattern" patented technology

The invention relates to the technical field of genomics, in particular to a gene detection data cleaning method and system based on artificial intelligence. The method comprises the following steps: acquiring gene detection data, and performing preliminary filtering on the gene detection data based on removal of low-quality reads to obtain preliminary filtering data; performing linker sequence identification and removal on the preliminarily filtered data, and performing sequencing error correction to generate gene sequence data; performing GC content analysis and repetitive sequence identification on the gene sequence data to obtain sequence feature data; performing dynamic quality threshold adjustment according to the sequence feature data, and performing secondary filtering on the gene sequence data to generate high-quality sequence data; and performing sequence pattern recognition on the high-quality sequence data to obtain sequence pattern feature data. According to the method, deep analysis is carried out by utilizing sequence characteristics such as GC content and nucleotide distribution, and the recognition capability on sequence complexity and variation modes is improved.

The invention provides a data flow monitoring method and system based on a large model, and the method comprises the steps: obtaining a data flow record set generated by a to-be-monitored system in a continuous operation period, carrying out the correlation path construction of the data flow record set, generating a data flow topological graph containing a node interaction relation and a time sequence dependency relation, and carrying out the correlation path construction of the data flow record set; calling a pre-trained circulation behavior analysis large model to perform node sequence pattern recognition on the data circulation topological graph, and generating behavior abnormal confidence and abnormal pattern labels of each node in the data circulation topological graph; and according to the abnormal behavior confidence and the abnormal mode label, screening an abnormal interaction node cluster in the data flow topological graph. According to the method, relevance between abnormal nodes and time sequence relevance are considered, missing detection or false detection is avoided, and the reliability of the monitoring effect is improved.

The invention discloses an abnormal data prediction method and system based on a dynamic threshold value and pattern mining, and the method comprises the steps: carrying out the mining of an association rule and a sequence pattern of historical monitoring data, recognizing the association relationship among a plurality of monitoring indexes under different data collection sources, and extracting a frequently appearing sequence pattern; setting a dynamic threshold range based on the mean value and the standard deviation of the historical monitoring data; dynamically adjusting the threshold range according to the real-time operation state and the change trend of the historical monitoring data; and through matching the real-time data with the historical mode, predicting a potential anomaly, and giving out an early warning. According to the method and the system, through adaptive dynamic threshold and abnormal mode mining, intelligent prediction and early warning of anomalies in a data acquisition process are realized, and post-processing is converted into beforehand early warning; through real-time monitoring and model analysis, the system can detect potential anomalies in advance, optimize data quality, improve operation and maintenance efficiency and reduce manual intervention.

The invention discloses a dynamic multivariate time sequence-oriented infrastructure load prediction method, which comprises the following steps of: predicting an infrastructure load by adopting a trained prediction model, segmenting acquired multivariate time sequence data, querying based on time steps and nodes of input segmented data to obtain a space-time pattern vector, and predicting the load of the infrastructure by adopting a dynamic multivariate time sequence-oriented infrastructure load prediction model. A sequence mode vector is obtained based on input segmented data query, the space-time mode vector and the sequence mode vector are spliced into a dual-mode vector output by a dual-mode pool module, and then in a dynamic MTS encoder, the dual-mode vector and segment embedding representation are spliced to obtain a final feature vector; and finally, inputting the final feature vector into a multi-scale decoder for decoding to obtain a prediction result. Aiming at challenges such as data imbalance, data missing and space-time dependence erosion, the accuracy of a prediction result is improved under the condition that repeated training is not needed, repeated training needed during topological change is avoided, and the robustness, prediction precision and calculation efficiency of the model are improved.

An attack path reasoning method based on attack technique and tactics scoring comprises the following steps: (1) malicious behavior traceability graph construction: constructing an original traceability graph based on system log data, detecting abnormal nodes, identifying attack techniques and tactics of the abnormal nodes, and retaining the abnormal nodes to form a malicious behavior traceability graph; (2) extracting an attack technology sequence mode: extracting an attack technology sequence from the threat intelligence, and constructing an attack technology sequence mode tree; and (3) scoring the edges of the malicious behavior traceability graph: scoring the edges in the malicious behavior traceability graph by comprehensively considering attack tactics and technologies. And (4) attack path reasoning based on the malicious behavior traceability graph: sampling of candidate attack paths is realized based on edge scores, scores of the candidate attack paths are calculated in combination with node feature information, and screening of the attack paths is realized. According to the method, attack path reasoning is carried out by comprehensively considering an execution sequence mode of attack techniques and tactics, so that more accurate attack path reasoning is realized; and the problem of alarm fatigue of the existing threat detection system is reduced.

The invention relates to the technical field of data processing, and discloses a wind turbine generator performance simulation analysis method and device based on time sequence analysis. The method comprises the following steps: performing high-frequency sampling and self-adaptive filtering on pre-acquired operation data of a wind turbine generator to obtain a denoised multi-dimensional time sequence data set; performing wavelet packet decomposition and Hilbert-Huang transform on the multi-dimensional time sequence data set to obtain a time-frequency domain feature matrix; performing nonlinear manifold learning and sparse representation on the time-frequency domain feature matrix to obtain a low-dimensional feature vector; performing kernel principal component analysis and information entropy calculation on the low-dimensional feature vector to obtain a quantitative index of the performance of the wind turbine generator; performing dynamic time warping and sequence pattern mining on the quantitative indexes to obtain abnormal working condition characteristics of the wind turbine generator; and performing fuzzy rough set analysis and fusion processing on the abnormal working condition features to obtain performance evaluation data of the wind turbine generator. The efficiency and accuracy of wind turbine generator performance simulation analysis based on time sequence analysis are improved.

The invention provides a commodity sales volume prediction method and system based on artificial intelligence, and relates to the field of data processing and intelligent prediction.The method comprises the steps that hot topic data are obtained, keyword tags in the hot topic data are extracted to be matched with commodity class tags of an e-commerce platform, and a hot associated commodity list is generated; and dynamic modeling of a user interest conversion path is combined, a user behavior sequence mode is analyzed, and a commodity popularity value of a real purchase intention is generated. The method comprises the following steps: carrying out time sequence sampling on popularity values, setting differential popularity threshold values based on commodity category features, dynamically identifying potential hot commodities, and calculating a stepped sales prediction value in future time in combination with a historical popularity value change curve and a sales conversion rate. In this way, the real-time performance and accuracy of sales volume prediction can be improved, the e-commerce platform can find potential hot commodities in time, inventory management is optimized, operation efficiency is improved, and prediction requirements in a complex and changeable market environment are met.

The invention discloses a low-altitude security event automatic classification and response method based on a cloud computing platform, and the method comprises the following steps: S1, collecting monitoring data in a low-altitude region, and carrying out the preprocessing of the monitoring data; s2, constructing a heterogeneous airspace behavior map, and extracting an evolution path and a behavior sequence pattern between nodes; s3, inputting the feature embedding matrix into a double-branch classification network, and respectively executing event type identification and event level evaluation; s4, constructing an event index, and designing a multi-objective strategy function; s5, the response instruction set is issued to the multi-source processing terminal through the cloud computing platform, and sealing control, interference, tracking or alarm operation is executed; and S6, performing incremental training on the graph convolutional neural network and the double-branch classification network, continuously updating network parameters and optimizing a multi-objective strategy function. According to the invention, high-precision identification and rapid intelligent response of low-altitude security events are realized, and the automation level and emergency disposal efficiency of airspace management and control are remarkably improved.

The invention relates to the field of artificial intelligence, and discloses a drilling fluid performance analysis and early warning method fusing a knowledge base and mode mining, which comprises the following steps: acquiring time sequence historical data such as drilling fluid density, funnel viscosity, plastic viscosity and sand content, and constructing a knowledge base module based on the acquired data; and mining a strong association rule of drilling fluid performance change by using a sequence pattern mining module of prefix projection. And finally, decomposing the data into a trend item, a periodic item and a random item through a time sequence decomposition technology, constructing an analysis and early warning module in combination with a long-short-term memory network, and sending out an early warning signal in time when monitoring that the performance parameters deviate from a normal mode or predicting that the performance is about to be abnormal. According to the method, a new effective method is provided for drilling fluid performance analysis and pollution early warning, the knowledge base and pattern mining are fused, the time sequence decomposition technology and time sequence long-term prediction are utilized, and the problems of overfitting phenomenon and abnormal value and noise tolerance are well solved.

The invention belongs to the technical field of data mining, and particularly relates to a commodity recommendation method and system based on efficient repeated negative sequence pattern mining. According to the method, the condition that negative conversion operation must be carried out in an item set range is not limited any more, conversion operation is allowed to be carried out on each commodity element on finer granularity, and therefore more repeated negative sequence candidate modes are generated; when the utility value of the repeated negative sequence candidate mode is calculated by the existing HUNSPM algorithm, the utility value of a negative commodity element is set to be 0, and the utility value of the negative sequence candidate mode is obtained by only considering the utility value based on a positive commodity element; when the utility value of the repeated negative sequence candidate mode is calculated based on the commodity element influence rate, the utility value of the non-overlapping positive sequence candidate mode is utilized, and the commodity element influence rate corresponding to the negative commodity element obtained through negative conversion processing is considered by utilizing IDSE (Pi, Si). The recommended commodities can better meet the requirements of customers.

The present invention belongs to the technical field of data mining, and particularly relates to a commodity recommendation method and system based on mining of high-utility repeated negative sequence patterns. The present invention no longer restricts that the negative transformation operation must occur within the itemset range, but allows the transformation operation to be performed on each commodity element at a finer granularity, thereby generating more repeated negative sequence candidate patterns; moreover, when the existing HUNSPM algorithm calculates the utility value of the repeated negative sequence candidate pattern, the utility value of the negative commodity element is set to 0, and only the utility value based on the positive commodity element is considered to obtain the utility value of the negative sequence candidate pattern; while when the present application calculates the utility value of the repeated negative sequence candidate pattern based on the commodity element influence rate, the present application not only utilizes the utility value of the non-overlapping positive sequence candidate pattern, but also utilizes IDSE(P i ,S i ) and considers the commodity element influence rate corresponding to the negative commodity element obtained by the negative transformation process. The commodity recommended by the present application can better meet the customer requirements.

The invention discloses a commodity recommendation method and system based on target sequence pattern mining, and belongs to the technical field of data mining. According to the method, a user shopping behavior sequence is coded to obtain a coded shopping sequence, and sequence candidate modes with the length of 2m + 1 are mined from the coded shopping sequence; deleting the modes which do not contain the target query sequence, taking the remaining modes as target sequence modes, and calculating the support degree of the target sequence modes; in the process of mining a sequence candidate mode with the length of 2m + 1 for a coding shopping sequence, when a target query sequence is a negative target query sequence, matching between a current prefix sequence and a precursor positive partner of a negative item and between the current prefix sequence and a subsequent positive partner of the negative item is realized under the guidance of the negative target query sequence. The method is short in time consumption for mining the target sequence pattern.

The invention discloses a chronic obstructive pulmonary disease patient screening system for the respiratory medicine department, particularly relates to the field of computer-aided diagnosis, and is used for solving the technical problems that an existing screening method is low in screening efficiency, low in patient adaptability and difficult to popularize on a large scale in a basic level. According to the system, a standardized patient medical event sequence is constructed by obtaining a target population medical record containing medicine distribution and disease diagnosis codes; identifying a characteristic medication and diagnosis mode before definite diagnosis of the chronic obstructive pulmonary disease based on a sequence pattern mining technology; constructing a medical knowledge graph by using the modes, obtaining a feature embedding vector representing a chronic obstructive pulmonary disease medical trajectory through graph neural network learning, and establishing a high-risk digital portrait; inputting the medical sequence of the person to be screened into a feature extraction model based on same map training, and generating an individualized risk probability prediction value by calculating the matching degree of the medical sequence and the digital portrait; and a screening result report is automatically generated, so that efficient and noninvasive early risk screening is realized.

The invention discloses a time sequence prediction method and system based on large language model semantic prompt enhancement. The adopted model comprises a semantic prompt enhancement module, a pre-training large language model and a self-adaptive mapping network; the semantic prompt enhancement module maps a vocabulary of the large language model into a sparse subset, and encodes the sparse subset of the vocabulary into a key matrix and a value matrix; embedding the time sequence into a vector code to form a query matrix, calculating the sparsity similarity between the query vector and a key matrix, and selecting the query vector to form a sparse query matrix; calculating sparse attention, and splicing the sparse attention with the time sequence embedded vector to obtain an enhanced time sequence embedded vector; the pre-trained large language model is used for embedding a vector according to the enhanced time sequence to obtain a time sequence preliminary prediction representation; and the adaptive mapping network adopts a B-spline function to carry out nonlinear mapping on the preliminary prediction representation of the time sequence. The modeling capability of the model on long-term dependence and complex sequence modes is enhanced, and the prediction precision is improved.

The invention belongs to the technical field of knowledge maps, and discloses a next interest point recommendation method based on a cross-regional city space knowledge map. According to the method, on the basis of the geographic space knowledge graph and the user preference knowledge graph, a cross-regional relationship modeling mechanism is introduced, the spatial proximity and the reachability between the regions are comprehensively considered, and more comprehensive description of the cross-regional behavior of the user is realized. According to the method, a recommendation framework combining a geographic module and a sequence module is designed, spatial dependence between interest points can be captured, a high-order sequence mode in user sign-in behaviors can be mined, and recommendation accuracy and diversity are improved. Geographic representation and sequence representation are fused through a consistency learning framework, the robustness and generalization ability of the model can be enhanced, and the model can still keep stable performance in sparse data and cold start scenes. According to the method, the defects of insufficient region boundary perception, recommendation result centralization, poor cross-region prediction adaptability and the like of an existing method are overcome.

The invention is suitable for the technical field of computers, and provides a game churn user recall method, which comprises the following steps: analyzing a behavior sequence mode before user churn, and determining a churn type of a churn user in combination with game event data before user churn; based on the churn type, making a personalized touch strategy for the churn user for sending periodic touch content to the churn user; whether the lost user triggers the in-game backflow activity or not is judged, a personalized backflow gift bag is configured for the backflow user, the recall efficiency and effect of the lost user can be improved, and powerful technical support is provided for long-line operation of the game.

The invention provides an Internet of Things equipment identification method and device, and belongs to the technical field of Internet of Things equipment identification, and the method comprises the steps: extracting an abnormal feature and a length feature of an Internet of Things equipment data packet, and constructing a multivariable time sequence feature based on the abnormal feature and the length feature; inputting the multivariable time sequence characteristics into a completely trained equipment identification model to obtain an identification result of the Internet of Things equipment; wherein the equipment identification model comprises a first Inception module, a first residual structure, a second Inception module, a second residual structure, a global average pooling layer and a full connection layer. According to the invention, the abnormal feature which can describe the influence of the network environment and the encrypted traffic on the traffic length sequence pattern of the data packet of the Internet of Things equipment is used as the input of the equipment identification model, so that the identification accuracy of the Internet of Things equipment can be improved.

The invention belongs to the technical field of Internet of Things. The invention provides a dynamic graph neural network link prediction method fusing multi-modal motifs. According to the embodiment of the invention, multi-source heterogeneous information such as structures, semantics, behaviors and the like is adaptively fused through a node-level modal attention mechanism, the problem that single-modal information is insufficient or unreliable is solved, and the discrimination ability of node representation is enhanced. And first-order adjacency, high-order adjacency and multiple motif modes are systematically fused, so that the model can simultaneously describe a local microcosmic mode and a global macroscopic topology, and the understanding of a complex network structure is more comprehensive. Spatio-temporal joint modeling is coordinated with time sequence attention through graph convolution, short-term mutation and long-term trend of network evolution are captured at the same time, and a dynamic link rule is accurately described. Through modal attention redistribution, time sequence consistency constraint and entropy regularization, modal missing and noise interference are effectively resisted, and prediction fluctuation is reduced.

The invention discloses a private cloud APT attack real-time detection method based on a fusion traceability graph, and the method comprises the steps: generating a current fusion traceability graph through the current log data of a private cloud platform and a low-level behavior sequence pattern mined from historical log data, extracting the features of the current fusion traceability graph, and inputting the features into a GRU model of a time sequence dynamic threshold value, enhancing the output of the GRU model by adopting a multi-head attention mechanism to obtain enhanced feature expression; calculating an abnormal score based on the enhanced feature expression and a plurality of normal system behavior clusters of the private cloud platform, and determining whether a behavior corresponding to the enhanced feature expression is abnormal or not according to a size relationship between the abnormal score and a preset threshold value; the plurality of normal system behavior clusters are obtained by generating a historical fusion traceability graph by using historical log data, extracting enhanced feature expressions of the historical fusion traceability graph, and clustering the enhanced feature expressions of the historical fusion traceability graph. According to the method, APT attack detection can be efficiently and accurately realized.

The embodiment of the invention discloses an SPM-LSTM-based load prediction method and device, and a medium, belongs to the technical field of micro-grids, and solves the problem of low micro-grid load prediction accuracy in the prior art. Comprising the following steps: constructing a data set containing load time sequence data based on obtained historical load data and historical meteorological data, and dividing the data set into a plurality of data subsets based on a time sequence time period; taking the plurality of data subsets as the input of the SPM, and determining a related sequence mode through the current data corresponding to the micro-grid and the SPM; wherein the correlation sequence mode is used for representing an incidence relation between the load data and the meteorological data; determining a reference data subset in the plurality of data subsets based on the correlation sequence pattern; selecting a reference LSTM model from a preset LSTM model set according to a time sequence time period corresponding to the reference data subset; and performing load prediction on the microgrid based on the reference data subset and the reference LSTM model.

The invention discloses a centralized management method and management system for enterprise-level work orders of a power supply system, and relates to the technical field of operation and maintenance management of the power supply system, and the method comprises the steps: collecting and preprocessing work order source data in the power supply system; constructing a work order association graph based on the work order source data, calculating a work order association degree, and forming a work order group; performing time sequence characteristic analysis on the work order group, identifying a periodic fault mode in the work order group by using a sequence mode mining algorithm, and generating a characteristic mark chain; training a dual-channel convolutional recurrent neural network by using the feature mark chain, establishing a fault prediction engine, and outputting a work order priority score table; and allocating a maintenance task instruction based on the work order priority score table, executing a maintenance task, and returning maintenance completion information to the work order association graph to complete work order whole process tracking. According to the method, the quantitative analysis of the association relationship between the work orders is realized by constructing the work order association map and combining the physical distance coefficient and the power supply path overlap ratio.

The invention discloses a commodity real-time recommendation method and system capable of efficiently mining by using a repeated negative sequence pattern, and relates to the technical field of data mining. According to the method, a prefix tree structure is designed and used for storing and maintaining an efficient repeated positive sequence mode, and efficient and dynamic management of data is achieved; the invention further provides a new efficient repeated negative sequence candidate mode utility value calculation mode, specifically, for any efficient repeated negative sequence candidate mode, the utility value of the efficient repeated positive sequence mode of the efficient repeated positive sequence mode matched with the efficient repeated negative sequence candidate mode is accumulated, and the effective value of the efficient repeated negative sequence candidate mode is calculated. And obtaining the utility value of the efficient repeated negative sequence candidate mode. According to the method, the accuracy of efficient repeated negative sequence pattern mining can be effectively improved, and commodities recommended by an application layer based on the more accurate efficient repeated negative sequence pattern obtained through mining can better meet the requirements of users.

The invention provides a vehicle fault early warning method, electronic equipment and a storage medium, and relates to the technical field of vehicle fault diagnosis and prediction. The method comprises the following steps: acquiring multi-source time sequence historical data in a vehicle running process within a first preset time period; discretizing the multi-source time sequence historical data based on the first preset time period to obtain a historical event sequence corresponding to each first preset time period; processing the historical event sequence by adopting a preset sequence pattern mining algorithm to obtain an event sequence pattern; determining a corresponding prediction rule based on the event sequence mode; performing monitoring matching on the multi-source time sequence real-time data and the prediction rule in the vehicle running process; and when the prediction rule is successfully matched with the multi-source time sequence real-time data, triggering target fault early warning of the vehicle. Through multi-source time sequence data mining and real-time matching, the limitation of isolated recording DTC is broken, early warning is triggered when matching succeeds, the passive situation of recording after a fault is thoroughly reversed, and beforehand early warning is achieved.

The invention relates to the technical field of intelligent water transportation information, in particular to an intelligent water transportation standard system construction method based on a four-dimensional level model. The method comprises the following steps: learning decision weights of different development value orientations according to idea features and execution features of a case, and determining a target development value of an actual water transportation service at a value dimension level; a spatial semantic tree is constructed by adopting a semantic layering mode, and traversal of the tree is driven by a target development value, so that matching of key element-scene requirements under a spatial dimension is realized. The method comprises the following steps of: partitioning a whole-cycle flow of a service, and performing time-dimension sequence pattern analysis on standard events at each stage to obtain corresponding time sequence construction requirements; and carrying out analysis and decision-making on operation paths among different elements from a logic dimension level to form a target implementation path and a support strategy. According to the method, a full-factor, full-period and full-scene intelligent water transportation standardization system can be constructed, and reliable multi-dimensional standard guidance is effectively provided for water transportation engineering construction.

A method for performing an ultrasound examination includes obtaining sequences of button pushes performed by a user via a control interface (125) during an exam workflow (S212); differentiating components of the exam workflow based on the button pushing sequences (S213); performing sequence pattern mining of the button pushing sequences based on the differentiated components to extract user specific workflow trends, which include most frequently used button pushing sequences (S214); detecting probe motion of a transducer probe (160) during the ultrasound examination (S215); predicting strings of next button pushes using a predictive model based on at least one previous button push in the user specific workflow trends, where predicting the strings of the next button pushes is triggered by the detected probe motion (S216); and outputting macro buttons corresponding to the predicted strings of next button pushes on the control interface, wherein selection of the macro buttons executes the corresponding predicted strings of next button pushes (S217).

The invention provides a group target trend prediction method and device based on sequence pattern mining, relates to the technical field of remote sensing target intelligent interpretation, and aims to solve the technical problems of single data source, weak multi-source data relevance, difficulty in accurately predicting cluster target trend and the like in the prior art. The method comprises the following steps: preprocessing multi-source data of a target group, and extracting time sequence data related to a navigation track and a parking area of the target group; according to the time sequence data, extracting a target group trajectory sequence pattern with the support degree higher than a support degree threshold value by using a sequence pattern mining model; based on the time-space correlation characteristics of each target trajectory in the target group trajectory sequence mode, a time sequence event atlas is generated, and the time sequence event atlas represents the behavior evolution path and state conversion relation of the group target at different time points and space points; and predicting the behavior trend of the group target by using a pre-trained long-short-term memory network model according to the time sequence event graph.

The invention provides a security protection method, system and device based on an API identity portrait and a medium, and belongs to the technical field of network security. The method comprises the steps of obtaining static attributes of a target API, including a data operation type, a transaction integrity requirement and a sensitive field, and determining a basic risk score B according to a preset coefficient; historical request logs are collected and analyzed, traffic baselines at different time points are determined, and the current traffic and the baselines are compared in real time to obtain behavior risk scores A; the method comprises the following steps: acquiring a call log, mining an API co-occurrence set by using FP-Growth, constructing a high-frequency sequence pattern through a PrefixSpan algorithm, and comparing a current session request to obtain a relation risk score N; summarizing attack frequencies and types in a preset time period, and obtaining an attack risk score R in combination with a historical statistical value; a total risk score W is determined based on the above scores, and whether to trigger detection is determined with reference to a preset threshold.

The application provides an object association recognition method and device and a storage medium, relates to the field of artificial intelligence, and the method comprises the following steps: obtaining target attribute features and target operation sequence data of a to-be-recognized object group; performing sequence pattern matching on the target operation sequence data based on a preset sequence pattern library to obtain a target operation sequence pattern of the to-be-recognized object group; performing sequence correlation processing on target operation features corresponding to the target operation sequence data by using a sequence correlation mining network of a preset relationship recognition model to obtain target sequence correlation features of the to-be-recognized object group; performing association relationship recognition on the to-be-recognized object group based on the target attribute features, the target operation features, the target operation sequence pattern and the target sequence correlation features by using an object group classification network of the preset relationship recognition model to obtain an association relationship category of the to-be-recognized object group. The application can effectively improve the generalization ability and timeliness of object association recognition.

The embodiment of the invention discloses an AI failure mode dynamic diagnosis method and system based on a sequence mode and function data analysis, which realize the crossing of AI interaction failure from static type identification to dynamic process diagnosis and improve the depth and timeliness of failure analysis. Accurate and explainable data support is provided for optimization of an artificial intelligence system. Through a sequence pattern mining and function data analysis technology, a failure pattern generated in an interaction process of a user and an artificial intelligence model (such as a large language model, a generative AI or an AI intelligent agent) is identified, analyzed and dynamically tracked, and the influence of the failure pattern is analyzed.

The present invention belongs to the technical field of message type recognition, and particularly relates to a message type recognition method and device based on data mining. The method includes first using a continuous sequence pattern algorithm to generate frequent continuous subsequences for a message sequence; then generating position-related candidate keyword fields on the selected frequent continuous subsequences through a key continuous sequence pattern algorithm; secondly calculating the probability of a candidate keyword field becoming a keyword based on a factor graph model; and finally selecting the candidate keyword field with the maximum probability as the keyword to determine the message type. The present invention uses data mining to quickly determine candidate keyword fields and improves the probability constraint relationship, and can accurately identify keywords and then determine the message type in a relatively short time.