136 results about "Degree distribution" patented technology

The invention discloses a bridge crack feature fine quantification method and system based on deep learning, and relates to the technical field of bridge detection, and the method comprises the steps: carrying out the preprocessing of bridge surface image data; inputting the preprocessed image data into a U-Net network fused with a multi-scale channel space attention module for semantic segmentation and outputting a crack segmentation mask graph; performing morphological refinement operation to obtain fracture skeleton line data, and calculating a trend angle and curvature distribution; measuring the crack width along the normal direction of the skeleton point to generate a width distribution thermodynamic diagram; accumulating skeleton point intervals to calculate the total length and identifying branch features; according to the method, the segmentation IoU reaches 85% or above, the width precision is superior to 0.05 mm, and the method has the crack development trend prediction capability.

The invention discloses a multi-dimensional distribution offset adaptive threshold detection method and system, and the method comprises the steps: obtaining real-time data, extracting a multi-dimensional statistical feature, and obtaining a feature vector; based on historical normal data, using an expectation maximization algorithm to train a Gaussian mixture model, and determining parameters to obtain a normal distribution model; inputting the feature vector into the model, and calculating a probability value of the feature vector belonging to normal distribution as a first offset judgment index; based on the real-time data distribution of a plurality of detection objects in the same group, the distribution difference of any two objects is calculated by using a Wasserstein distance, and the similarity between the objects is obtained; and constructing a similarity network and calculating connectivity as a second offset judgment index. Setting a fixed-length sliding window, dynamically updating two indexes in the window, and obtaining a first self-adaptive threshold value and a second self-adaptive threshold value; and when any index is lower than a corresponding threshold value, determining distribution offset and giving an alarm, and updating model parameters in real time by using an incremental expectation maximization algorithm. According to the invention, accurate detection and intelligent analysis of data distribution offset are realized.

The invention discloses a context poisoning attack defense method oriented to an RAG application, and relates to the technical field of RAG. the method comprises the following steps: inputting a target query statement, and retrieving the target query statement to obtain multiple pieces of context information; taking representative sentences in the retrieved context information, and identifying and filtering potential malicious template clusters; the big language model gives all candidate answers according to existing context information, the logarithmic probability of all contexts to different candidate answers is calculated, and after the influence of parameter knowledge of the big language model is removed from the logarithmic probability, the support degree of all contexts to different candidate answers is obtained; the whole logarithmic probability vector is used as a support degree distribution condition of the context to the candidate answers; identifying a single piece of harmful information from the support degree distribution condition of the context to the candidate answers through a logistic regression model so as to filter wrong answers; according to the attack defense method provided by the invention, centralized injection of multiple malicious texts and sparse injection of a small number of malicious texts can be defended.

The invention provides a fault root cause analysis and disposal scheme generation method based on a large model, and relates to the technical field of artificial intelligence and fault diagnosis and operation and maintenance, and the method comprises the steps: generating a fault feature vector sequence fusing spatio-temporal topological semantics; inputting the fault feature vector sequence into a pre-trained large model to construct a fault propagation chain, and judging whether the system has a fault or not based on an abnormal strength threshold value; if the fault exists, abnormal nodes are identified through a large model, and a correlation score of each abnormal node is calculated based on a time sequence change feature and a causal attention mechanism; calculating to obtain responsibility degree distribution of each abnormal node through a large model according to the node topological relation and the correlation score; and adjusting the responsibility degree value according to the responsibility degree distribution and the depth of the abnormal node in the call chain, and outputting a fault root cause analysis and disposal scheme recommendation result according to a responsibility degree threshold. Intellectualization of root cause analysis is achieved, a targeted recovery operation instruction is provided, and the fault recovery time is greatly shortened.

The invention discloses a multi-scene risk perception method based on a knowledge graph and deep learning, and the method comprises the following steps: collecting multi-source heterogeneous data, carrying out the unified standardization of the multi-source heterogeneous data into a multi-modal data vector, converting the data vector into polar coordinate representation with direction and amplitude characteristics, and carrying out the inter-modal fusion through a polar coordinate transformation network; the fusion result is matched with a preset risk knowledge graph, and a dynamic graph structure containing risk nodes and edge weights is constructed; by introducing a reverse stochastic differential equation, the state of risk evolution along with time is simulated; and analyzing the risk density distribution in different directions, generating a dynamic vector for controlling risk reasoning, dynamically activating a risk channel in an asynchronous mode, and comprehensively outputting a unified multi-scene risk perception map. According to the invention, continuous and interpretable identification and prediction of the risk state in a complex environment are realized.

The invention discloses a robot group detection scheduling method and system, and relates to disaster monitoring: scene reconstruction is carried out according to collected real-time sensing information, a three-dimensional digital model of a disaster scene is generated, and a discrete monitoring grid based on sampling points is established; constructing a disaster risk assessment model on the discrete monitoring points, and for gas and fire, calculating the risk value of each monitoring point by using actual measurement data to form a discrete risk distribution map; calculating the space change trend of the risk degree by adopting an adjacent point difference method, and calculating a directed gradient through the risk degree difference value and the distance of adjacent monitoring points; when the risk degrees of any two disasters are increased at the same time, marking the risk points as potential coupling risk points; identifying a propagation path of the risk degree based on connectivity analysis of the monitoring network; and scheduling the robot group by adopting a risk avoiding path planning algorithm. According to the method, the spatial distribution and evolution trend of the disaster field are accurately reconstructed through local gradient estimation under the sparse sampling condition.

The invention relates to the technical field of constructional engineering quality detection, and discloses a method for testing the apparent quality of mirror-surface bare concrete. The method comprises the following steps: acquiring a multiband multispectral image of the surface of a component; through image fusion and feature decoupling, an independent specular glossiness distribution diagram, a microscopic fluctuation topological diagram and a chromaticity coordinate mapping diagram are obtained; registering and associating the three images to construct a three-dimensional total element apparent characteristic field; performing gridding scanning on the characteristic field to extract glossiness uniformity, texture roughness and chromaticity stability indexes of each grid; inputting a pre-trained apparent quality analysis network to output a local quality evaluation value and a defect type label of each grid; and integrating and generating an overall quality index and generating a repair guidance map according to the defects. According to the method, multi-dimensional, full-field and intelligent fine detection of the apparent quality is realized, the gloss, the texture and the color can be independently and quantitatively analyzed, and the defect can be accurately positioned.

The embodiment of the invention provides a multi-account concurrent processing management system and a multi-account concurrent processing management method, which are applied to the technical field of multi-account high-concurrency business processing, and are used for analyzing a plurality of received account operation requests into a micro-operation set, packaging micro-operations into an operation object identifier, an operation type and a dependency relationship, and storing the operation object identifier, the operation type and the dependency relationship. Comprising resource dependence, granularity dependence and business logic dependence; determining a lock execution path based on the operation object identifier, the operation type and the dependency relationship; constructing a dependency tensor model, performing conflict density mapping to form conflict density distribution, and performing tensor conflict operation on the microoperation in combination with a lock execution path to obtain a conflict intensity value; a high-conflict area is identified based on conflict density distribution, a potential deadlock graph is constructed, a micro-operation directed acyclic graph is generated, nodes with zero in-degree and out-degree are identified to form a first execution sequence, a second execution sequence is constructed according to residual micro-operations, a micro-operation execution sequence is formed through splicing, and efficient parallel scheduling and throughput capacity improvement are achieved.

The invention provides an intelligent work order automatic classification system, and relates to the technical field of work order management. According to the system, work order texts are analyzed through an LLM technology, core business phenomena are identified, auxiliary descriptions are eliminated, and an initial phenomenon mark set is generated. And further establishing a cross-sentence association analysis mechanism, dynamically inserting context associators with directivity between discrete phenomena based on syntactic dependency and semantic role labeling, and implementing dynamic weight assignment based on context proximity. The system generates a causal rule set with version identification, and a multi-decision trigger unit processes weighted phenomenon marks in parallel to generate a differentiated candidate root cause set. And the decision conflict quantization unit calculates rule matching degree distribution, and activates rule backtracking verification to ensure decision stability. And finally outputting a root cause classification work order entity with a rule version traceability identifier, thereby realizing automatic work order classification with high accuracy and stability.

The invention relates to a supply chain demand prediction and risk early warning method and system based on multi-source data. The method comprises the steps of collecting multi-source heterogeneous data and constructing a feature vector set; constructing a supply chain network topology structure; inputting the feature vector set and the supply chain network topology structure into a risk propagation model to obtain a risk space-time distribution diagram; inputting the feature vector set and the risk space-time distribution diagram into a demand prediction model to obtain collaborative prediction result information; based on a supply chain network topology structure, generating a node criticality distribution map, comparing, identifying high-risk nodes, obtaining risk level information, and generating graded risk early warning information; according to the method, the feature vector set and the supply chain network topology structure are constructed by integrating the multi-source heterogeneous data, combined optimization of dynamic risk simulation and demand prediction is realized by combining cooperative calculation of the risk propagation model and the demand prediction model, and the method has the advantages of improving prediction accuracy, dynamically identifying high-risk nodes and generating hierarchical early warning information.

The invention discloses an LDPC (Low Density Parity Check) code check matrix construction method, which is used for processing a low earth orbit satellite ground foundation zone, and comprises the following steps of: A1, determining degree distribution of a basis matrix under a high code rate by adopting an extrinsic information transfer graph EXIT, and realizing construction of the basis matrix under the constraint of optimal degree distribution through extrinsic information transfer P-EXIT based on a basic model graph; a2, based on the code length and the code rate of FEC coding of the DVB-S2 protocol, determining the size and the expansion factor of a basis matrix corresponding to each code rate; a3, determining a finite field and a generator based on an expansion factor, and constructing a linearly independent shift value row vector and cycle coefficient table; a4, selecting row vectors from the cyclic coefficient table to construct a cyclic shift matrix, and performing short ring identification and elimination on the cyclic shift matrix; and A5, performing matrix hashing on the basis of the basis matrix and the cyclic shift matrix to obtain an LDPC check matrix.

This invention discloses a method for constructing a natural-artificial water network model and analyzing its connectivity, including: collecting basic data; generalizing nodes and edges of the natural and artificial water networks and adding weights to the edges; constructing a natural-artificial water network model based on complex networks; constructing the degree distribution of structure and flow based on node degrees; and calculating the structural connectivity and flow connectivity of the natural-artificial water network based on the degree distribution. This invention combines complex network theory to model natural-artificial water networks and analyze their connectivity, solving the technical problem of lacking comprehensive modeling and connectivity assessment in existing water network analysis methods, and providing new analytical ideas and technical support for water resource management and water network structure optimization.

The present disclosure belongs to the technical field of network security, and particularly relates to a network anomaly monitoring method, device and program product based on traffic fingerprint learning, which comprises the following steps: extracting metadata irrelevant to encryption of a target network; the metadata comprises five-tuple, timestamp, length and protocol flag information of a data packet; the metadata is aggregated with a source IP address as a primary key to define a network entity; session feature extraction is performed on a session of the target network entity and other network entities; the extracted session features comprise timing features, operation sequence features and scale distribution features; the operation sequence features comprise Shannon entropy of a continuous request and response type pair; an action fingerprint of the target network entity is constructed according to the session features; and whether the target network and the target network entity are abnormal is judged according to a deviation degree of a current action fingerprint of the target network entity and a fingerprint baseline. The present disclosure can realize accurate identification of network anomalies.

The invention relates to the technical field of modern communication, and discloses a wireless network card power adaptive control method and system. The method comprises the following steps: acquiring real-time interference data, performing spectrum change according to the real-time interference data to obtain an interference intensity distribution diagram, performing interference correlation analysis and power optimization calculation according to the interference intensity distribution diagram to obtain a power configuration scheme, performing parameter updating according to the power configuration scheme to obtain a preliminary power correction result, and performing power optimization calculation according to the preliminary power correction result. The method comprises the steps of obtaining a preliminary power correction result, carrying out interference comparison evaluation according to the preliminary power correction result to obtain deviation information, carrying out power optimization according to the deviation information to obtain an optimized power configuration scheme, and carrying out synchronous configuration and index feedback analysis according to the optimized power configuration scheme to obtain a final regulation command set. According to the method, the problem that interference chain reaction is easily caused by equipment power adjustment can be solved, so that stable communication and interference suppression in a complex network environment are realized.

The present application relates to the field of engineering detection technology, in particular to a deep compaction degree nondestructive testing method and system for large-thickness water-stable base course. A multi-source heterogeneous dataset of a to-be-detected area is obtained using an integrated detection device; a construction process compliance index and a process uniformity index of the to-be-detected area are obtained; each data source in the multi-modal nondestructive testing data is deeply corrected to obtain a deep correction compaction degree estimation set of each data source; the deep correction compaction degree estimation set of each data source is input into an adaptive weight fusion model to obtain a preliminary compaction degree distribution map; a fusion compaction degree distribution map of the to-be-detected area and a dynamic confidence score of each spatial point are calculated; the compaction quality of the to-be-detected area is determined, and a detection strategy optimization instruction is generated for the area with a confidence score lower than a confidence threshold. The present application can realize high-precision, high-confidence and engineering-actual nondestructive testing of deep compaction degree of large-thickness water-stable base course.

The invention provides a connecting rod fatigue life evaluation method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining connecting rod fatigue life original data under multistage loads, constructing a structured data set, and training a deep neural network model according to the structured data set; predicting the life uncertainty of each load point in a preset target load range according to the trained network model, selecting the load point with the maximum life uncertainty to perform a fatigue test, and updating the network model according to a test result to obtain optimized S-N curve parameters and life distribution parameters; and based on the real-time load spectrum data, the optimized S-N curve parameters and the life distribution parameters, carrying out variable-amplitude load life calculation through a life distribution formula and a damage formula to obtain real-time fatigue life and damage contribution degree distribution under different preset survival rates. The accuracy of fatigue life evaluation of the connecting rod is improved, and accurate prediction and dynamic evaluation of the fatigue life under different survival rates are realized.

The invention relates to the technical field of physical field prediction, in particular to a physical field prediction method and device based on gradient recognition parameter tuning and a medium, and the method comprises the steps: determining a space-time computational domain based on a physical problem, and determining a control equation, an initial condition constraint and a boundary condition constraint of the physical problem; constructing a training sample data set; initializing parameters of the neural network, and pre-training the neural network; obtaining pre-trained neural network parameters, and identifying high-contribution neural blocks; on the basis of the high-contribution-degree nerve blocks, parameter fine tuning iteration is carried out on the weights of the combination points of the neural network, and a physical field prediction model is obtained; and inputting the space-time coordinates of a to-be-solved point into the physical field prediction model to obtain a physical field prediction value of the point. The adaptive algorithm framework and the staged training strategy provided by the method enable the model to adapt to different partial differential equations and gradient distribution characteristics, thereby reducing the use threshold of the physical information neural network.

A denoising method based on a multiscale distribution score for a point cloud includes: constructing a two-layer network model based on multiscale perturbation and point cloud distribution, where the two-layer network model includes a feature extraction module for extracting a feature of the point cloud and a displacement prediction module for predicting a displacement of a noise point; constructing a point cloud noise model for improving a denoising effect and retaining a sharp feature and avoiding reducing quality of point cloud data; extracting a global feature h by inputting the point cloud data into the feature extraction module; iteratively learning the displacement of the noise point by the displacement prediction module according to a feature obtained by the feature extraction unit; and defining a loss function of network training, and completing convergence under the condition that the loss function reaches a set threshold or a maximum number of iterations.

The application belongs to the technical field of target identification, and provides a dynamic target identification method and system based on a low-illumination environment. An illumination component and a reflection component are separated frame by frame through a Retinex decomposition network, and an illumination distribution map is generated based on the illumination component. The convolution kernel weight of an edge branch is modulated based on the illumination distribution map, and an edge feature map and a semantic feature map are extracted. A candidate region and a corresponding uncertainty distribution are generated based on the illumination component and the reflection component to determine a fusion weight and complete weighted fusion, thereby improving the integrity and recognition degree of the fusion feature map. Target detection and confidence determination of the candidate region are performed based on the fusion feature map, local secondary enhancement is performed on a low-confidence region, and a target region is output after inter-frame residual fusion based on the fusion feature map of adjacent frames, thereby improving the stability and environmental adaptation capability of cross-frame dynamic target identification.

This invention provides a refined design method for explosive charge quantity based on rock mineral particle structure, relating to the field of geotechnical engineering blasting technology. This invention maps the mineral composition and particle size data of rock samples to mineral hardness distribution characteristics and particle structure characteristics, respectively, and generates a high-dimensional feature vector through nonlinear encoding fusion. This vector is input into a graph neural network, and a rock structure feature relationship map is formed through node representation and attention mechanisms. Using this map as a condition, under the constraint of baseline explosive consumption, a deep generative network learns and outputs a differentiated charge distribution matrix corresponding to spatial location. After alignment and transformation with blasting hole mesh parameters, a charge allocation design map is generated. Finally, after blasting based on this design map, actual effect indicators are collected to construct a multi-objective optimization function. An optimization algorithm is used to adjust the parameters of the graph attention mechanism and the generative network in reverse according to the effect differences, achieving iterative optimization of the model.

The application discloses a natural gas load interval estimation method based on large model coupling working condition clustering, and belongs to the technical field of natural gas pipeline network operation optimization and artificial intelligence load prediction. The method extracts working condition semantic constraints by using a large model, and obtains fuzzy working condition clusters by combining historical operation data clustering; a natural gas load point prediction model is trained for each cluster, and residual probability density distribution is estimated; semantic and numerical weights are fused in real time, and the final point prediction value and the natural gas load prediction interval at the prediction time are obtained by dynamic weighting, which are taken as the natural gas load prediction result and output. The application introduces a large model into the natural gas working condition expression and clustering constraint construction process, and no longer uses the large model as a simple downstream feature generation tool, but solves the problem of interval estimation failure of the natural gas load under fuzzy working conditions such as holiday switching, peak-valley transition and extreme weather through deep coupling of the large model and working condition clustering, so that high-reliability dynamic prediction interval output under complex working conditions is realized.

This application relates to the field of smart property management and discloses a resource scheduling method for smart property management based on deep learning. The method includes: constructing a spatial topology influence map of the target property management area; spatially aligning the multi-source heterogeneous data streams according to the nodes of the spatial topology influence map and generating a sequence of node feature vectors corresponding to each node; inputting the node feature vector sequence into a spatiotemporal influence propagation prediction model; performing spatial feature aggregation on the node feature vector sequence based on the graph attention layer and the static topology constraint parameters in the spatiotemporal influence propagation prediction model to generate a spatiotemporal influence representation characterizing the degree of impact of sudden events on each functional area; predicting the resource demand heat value of each functional area within a preset time period based on the spatiotemporal influence representation to form a resource demand heat distribution; and generating a resource scheduling instruction set based on the resource demand heat distribution. This technical solution improves the user's property service experience.

The application discloses a kind of based on the data transmission method and system of splicing processor of multi-path data synchronization, belong to data transmission field.The application obtains the depth video data of multiple overlapping area coordinates and time stamp and real-time network load, first, overlapping area coordinates are used to detect adjacent frame boundary and extract depth difference to generate horizontal and vertical pixel displacement compensation parameters, then correct each video frame in pixel position and splice as panoramic depth video frame sequence.Based on time stamp, calculate the clock difference of each road, when exceeding synchronization threshold, corresponding frame sequence is implemented buffer time stamp incremental reordering.Subsequently, extract the depth gradient distribution features in frame, identify the continuous area of depth change rate exceeding threshold as high priority data block, combine network load and spatial density coefficient to calculate transmission weight, finally, realize the block priority transmission of panoramic video according to weight.The application can realize low-latency data transmission in complex network environment.

This invention discloses a method and apparatus for analyzing the degree distribution of business data based on graph theory. The method includes: acquiring business data to be analyzed and converting the business data into a network graph representation to obtain a business data network graph; calculating the degree of each node based on the business data network graph and drawing a business data degree distribution graph based on the degree of each node; obtaining network structure information based on the business data degree distribution graph and performing parameter fitting on the business data degree distribution graph based on the network structure information; correcting for missing edges on the fitted degree distribution graph to obtain the analysis results of the business data degree distribution; and then visualizing the analysis results of the business data degree distribution to complete the analysis of the business data degree distribution. This invention solves the technical problems of low accuracy and difficulty in analysis and visualization of the degree distribution of business data networks in the prior art.

The application relates to the technical field of intelligent auditing, and discloses a digital import and export intelligent auditing method based on a CV-NLP fusion algorithm, which comprises the following steps: performing image coordinate standardization processing on an image sequence of an import and export document package, generating a set of mark candidate blocks containing continuous and reliable weights by using semantic probability and prior attributes of a layout, calculating a layout level complexity parameter representing the fragmentation degree of the hierarchical structure in the document package according to the normalized area scale distribution of the block set, generating a multi-level resolution sampling weight distribution based on the parameter, and performing multi-scale identification and probability fusion on the candidate blocks by using the weight determined by the distribution, calculating a character structure fragmentation measurement value representing the structural fragmentation bias and a region positioning volatility measurement value representing the dispersion degree of the cross-page layout landing point in combination with the layout geometric attributes, and finally constructing a comprehensive risk assessment model to output an intelligent auditing result quantitatively representing the document level identity observability risk.

The application belongs to the technical field of communication system, and particularly relates to a method and system for deriving optimal degree distribution of regular LDPC code based on AWGN channel. The method comprises the following steps: S1, taking the maximization of code rate of the regular LDPC code as a criterion and taking the decoding success as a prerequisite, a target function with a constraint condition is constructed; S2, fixing the degree d v of the variable node, based on the fixed point analysis theory, a theoretical analytic expression of the degree d c of the check node in the LDPC code is derived, and an inverse function of the target function in the non-convex optimization problem is solved; S3, based on the mathematical set theory, the uncertainty in the inverse function solving problem in the step S2 is solved, and the optimal degree distribution of the LDPC code under the AWGN channel is obtained. The application has the characteristics of improving the channel coding performance in the communication system.

The invention provides a business scene recommendation method which can be applied to the technical field of artificial intelligence. The business scene recommendation method comprises the steps of obtaining a demand description text of a target business; matching degrees between the demand description text and m preset hot word banks are obtained, first matching degree distribution is obtained, the m preset hot word banks are set based on keywords of m service scenes, the m service scenes comprise n preset service scenes, m is an integer larger than 1, and n is an integer larger than or equal to 1; in response to the target service as the preset service scene, identifying the demand description text by using the pre-training language model to obtain second matching degree distribution, the second matching degree distribution being matching degree distribution of the target service and n preset service scenes; and obtaining a recommendation result based on the first matching degree distribution and the second matching degree distribution. The invention further provides a service scene recommendation device and equipment, a storage medium and a program product.

The invention discloses an SOME / IP service flow-oriented spatio-temporal behavior perception and dynamic shaping method, which comprises the steps of collecting SOME / IP message mirror images on a bypass of a vehicle-mounted network controller main control platform, analyzing and extracting spatio-temporal multi-dimensional features, constructing a service granularity flow sample set based on a sliding window, completing kernel density estimation through self-adaptive gridding discretization and frequency domain accelerated convolution, and obtaining a service granularity flow sample set. Service flow probability density distribution is deeply represented, and time-space communication behaviors of the service flow are quantified through comprehensive information entropy; on-line comparison is decided based on a self-adaptive double-entropy threshold value, and a hierarchical shaping strategy is accurately triggered; and driving the release end ECU to execute in real time through a standard interface, constructing closed-loop recursive feedback by fusing a bottom layer physical state to carry out dynamic correction, and setting a state deviation degree-based severe boundary constraint and deterministic failure rollback mechanism. According to the method, SOME / IP service-level precise management and control, low-delay shaping and vehicle-gauge-level safety guarantee are realized, a bottom-layer protocol stack does not need to be modified, and a data-driven dynamic traffic management normal form is constructed for a vehicle-mounted network.

The invention provides a fault-tolerant processing method and device for incomplete data, and the method comprises the steps: recognizing the incomplete data with errors and missing, and determining an abnormal data sample set containing errors and non-random missing in the incomplete data; determining error distribution and missing probability density distribution of data key feature dimensions in the abnormal data sample set; determining a dynamic association degree matrix of characteristic variables between the financial association data and the incomplete data, and determining an adaptive characteristic weight vector of the incomplete data according to the dynamic association degree matrix, the error distribution and the missing probability density distribution; determining reconstructed complete data of the incomplete data in combination with historical financial core data related to financial risk perception; and performing distributed fault-tolerant verification on the reconstructed complete data, performing homomorphic encryption processing on a fault-tolerant verification result, and transmitting the fault-tolerant verification result to a financial risk perception data storage end. By adopting the scheme of the invention, fault-tolerant processing can be carried out on the data in combination with errors and non-random missing of the data.

This invention discloses a method and system for promoting and selling rubber coating products, belonging to the field of computer-aided sales technology. The method includes: acquiring customer interaction data of rubber coating products during historical promotion periods to generate a historical attention distribution map; performing demand feature decomposition on the distribution map to obtain historical demand tendency vectors and establishing a demand evolution trend tensor; constructing a customer demand semantic space based on the demand evolution trend tensor, calculating the semantic clustering degree of historical transaction events to determine potential transaction feature vector clusters; extracting real-time demand features from real-time promotion feedback data to form real-time demand semantic vectors, calculating their semantic correlation with potential transaction feature vector clusters in the semantic space to obtain a real-time transaction potential index; and finally combining the historical demand tendency vectors and the demand evolution trend tensor to generate a transaction probability prediction function, predicting the real-time transaction probability, and generating promotion strategy adjustment parameters.