34 results about "Bayesian statistics" patented technology

The invention discloses an event prediction and early warning method based on Bayesian deep learning. The event prediction and early warning method mainly comprises the following two parts: constructing a Bayesian deep learning rule model based on event characteristics, and designing a prediction and early warning model on the basis of constructing the Bayesian deep learning rule model. According to the method, the Bayesian statistical method and the deep learning model are fused, the advantages of the Bayesian statistical method and the deep learning model can be fully utilized, the uncertainty of a prediction result can be expressed in a probability distribution form by combining priori knowledge and observation data, efficient modeling and prediction are performed on complex and nonlinear time sequence data, early warning of potential events is realized, and the prediction efficiency is improved. The accuracy and reliability of event prediction and early warning are improved.

The invention discloses a mulberry SNP (Single Nucleotide Polymorphism) marker mining and stress resistance character prediction method based on a Transform architecture, and belongs to the crossing field of biotechnology and artificial intelligence. According to the method, mulberry genome sequencing and stress resistance phenotype data are collected, high-credibility SNP is screened through quality filtering and Bayesian statistics, a Transform model containing multiple self-attention mechanisms is constructed after multi-dimensional features are extracted, early stop method training optimization is combined, stress resistance character prediction is achieved, and the key SNP contribution degree is analyzed through feature disturbance. According to the method, the problems of insufficient long-distance feature correlation capture, low prediction precision, poor model interpretability and the like in the traditional technology are effectively solved, the stress resistance character prediction precision and the SNP marker credibility are improved, and an interpretable efficient screening strategy is provided for mulberry molecular breeding.

A multi-field data analysis method based on a Bayesian information enhanced neural network comprises a BITRNN model, and a final optimization objective of the model combines expected cumulative rewards of reinforcement learning, an evidence lower bound (ELBO) of Bayesian reasoning and an exploration mechanism of information entropy. Through combination of Bayesian reasoning, information entropy and reinforcement learning, prediction and decision optimization in multi-field complex data are realized. The model provided by the invention can dynamically adapt to environmental changes and maintain efficient prediction performance in high-dimensional uncertain data. Through Bayesian reasoning, the model can effectively process parameter uncertainty; the introduction of the information entropy increases the exploratory performance of the strategy, and avoids falling into a local optimal solution. According to the method, theoretical advantages of Bayesian statistics, an information theory and reinforcement learning are fused, high-dimensional and high-uncertainty data can be effectively processed, the prediction precision and generalization ability of the model are improved, the method is suitable for various fields such as polymer material performance optimization, agricultural planting strategy making and financial investment decision making, and the method has wide application prospects. Wide application prospects and practical values are realized.

The invention relates to the technical field of link transmission optimization, in particular to an inter-satellite multipoint transmission and link dynamic optimization method based on Bayesian situation awareness. Comprising the following steps: accessing an interstellar communication network interface, collecting real-time link data including power, frequency and signal-to-noise ratio, recording the data transmission rate and delay of each link and the data packet loss rate, and generating real-time link state information. According to the method, the link state is monitored in real time, performance prediction is carried out through Bayesian statistics, potential delay and abnormity are effectively predicted, the transmission path is optimized through the Di jkstra algorithm, the method adapts to the constantly changing environment of interstellar communication, the transmission efficiency of data under the extreme condition is ensured, and the method has the advantages of being high in practicability and the like. The dynamic routing adjustment strategy is updated according to the real-time network load, the maximum efficiency of data transmission is ensured, the automatic network topology reconstruction further strengthens the self-healing capability of the system, and the shutdown time and potential performance degradation of the system are reduced through quick response.

The invention discloses an overwater oil spill and hazardous chemical substance leakage accident pollution source and pollutant distribution inversion method. The method comprises the following steps: obtaining hyperspectral remote sensing data at an accident site; establishing a mathematical model to describe the transmission and diffusion process of the pollutants in the water body; according to a Bayesian statistical inference method, the posterior probability distribution of the pollution source position and the pollutant concentration is updated in combination with the actually measured data; the pollution source position and the pollutant concentration are optimized through an optimization algorithm, so that the prediction result of the mathematical model is matched with the actually measured data as much as possible; and the reasonability and credibility of the inversion result are analyzed and evaluated, and are compared and verified with the actual situation, so that the scientificity and accuracy of the inversion result are ensured. According to the method, a Bayesian statistical inference and optimization method is used, inversion can be carried out more effectively by using measured data, accurate estimation of the position and the initial concentration of the pollution source is realized, and the data-driven method has higher flexibility and adaptability when processing complex environmental conditions and change conditions.

The present invention relates to a method for processing document data by splitting and embedding long documents and removing duplicate chunks. The method is configured to efficiently split long documents, remove chunks with high redundancy to leave only core chunks, and utilize them in question-answering systems or learning models. First, long documents are divided into chunks of a fixed size, and contextual disconnection is minimized by overlapping some tokens between adjacent chunks. Subsequently, each chunk is vectorized using an embedding model, and a novelty score is calculated using the similarity (redundancy) between vectors. This novelty score is compared with a threshold updated using Bayesian statistical techniques, and chunks with high redundancy are automatically removed. When only the selected chunks are finally connected and input into a Generative AI model, a free-form answer to a user question (Q) is generated, focusing on core information while excluding redundant content.

This application provides a business data prediction method based on multi-model fusion, applicable to the fields of artificial intelligence, big data, and fintech. The method includes: predicting incremental data for a target year using a Bayesian statistical model based on multi-source business data; obtaining incremental data for a target quarter using a seasonal time-series prediction model based on the multi-source business data; performing long-term decomposition prediction using the seasonal time-series prediction model and a multinomial model based on the incremental data of the target quarter to obtain initial prediction data for the target year and fluctuation data for the remaining quarters; wherein the target quarter and the remaining quarters constitute the target year; and using the incremental data of the target year as trend reference data, and based on the fluctuation data of the remaining quarters, correcting the initial prediction data for the target year to obtain the business data prediction result. This application also provides a business data prediction apparatus, device, storage medium, and program product based on multi-model fusion.

The present application discloses a method and system for evaluating the hydrogen-induced fatigue life of bolts based on the Bayesian method, which relates to the field of fatigue life evaluation. The method includes obtaining a fatigue constitutive model of the bolt at the material level; determining a basic fatigue characteristic curve according to the fatigue constitutive model of the bolt at the material level and material experimental data, and using the finite element analysis method to perform stress analysis on the actual structure of the bolt to determine the stress concentration coefficient; correcting the fatigue constitutive model at the material level according to the stress concentration coefficient and the hydrogen embrittlement correction function to obtain a corrected fatigue constitutive model; determining a non-linear cumulative damage model according to the corrected fatigue constitutive model by using the non-linear cumulative damage theory; optimizing the non-linear cumulative damage model by using the Bayesian statistical method according to the material experimental data. The present application can improve the accuracy and reliability of bolt fatigue life prediction.

The invention discloses a wind speed multi-mode integrated probability forecasting method based on dynamic classification and logarithmic normal distribution. The method comprises the following steps: dividing a forecasting area into a plurality of sub-areas; dynamically determining the optimal set quantile of each sub-region, and dividing the wind speed training samples of each sub-region into a low-wind-speed training sample set and a high-wind-speed training sample set by comparing the relationship between the predicted value corresponding to the optimal set quantile and the 95th percentile of the observed wind speed; on the basis of a Bayesian statistical framework, logarithmic normal distribution is adopted as a probability density function, the low-wind-speed training sample set and the high-wind-speed training sample set of each sub-region are modeled, and a differentiated wind speed probability forecasting model is constructed; and dynamically selecting the wind speed probability prediction model of the corresponding classification for each sub-region to perform wind speed prediction. According to the invention, the problem of insufficient adaptability of a traditional single model to regional differences is solved; meanwhile, objective classification of wind speed is realized, a differentiated parameter system is established, and the wind speed probability forecasting skill is improved.

This invention relates to a lane navigation method and system, a computer device, and a storage medium, comprising: receiving a forward image captured by an onboard camera within the current period; inputting the forward image into a pre-trained deep learning model for image recognition to obtain the speed (Speed_A) and brake light activation count (Count_A) of vehicle A in the current lane, and the speed (Speed_B) and brake light activation count (Count_B) of vehicle B in the adjacent lane; performing Bayesian statistics on the brake light activation count (Count_A) and brake light activation count (Count_B) to obtain the braking deviation probability of vehicle A or vehicle B at the next moment; determining whether to change lanes to the adjacent lane based on the relationship between the speeds (Speed_A and Speed_B) and the braking deviation probability of vehicle A or vehicle B at the next moment; and outputting corresponding navigation instructions based on the results of the above determination, thereby improving the accuracy of lane navigation.

The application discloses a complex system coupling coordination degree optimization method and device, and the method comprises the following steps: acquiring a multilevel coupling parameter set in a complex system; screening a to-be-optimized parameter variable from the multilevel coupling parameter set to obtain a to-be-optimized parameter variable set; constructing a Bayesian statistical model; taking the Bayesian statistical model as a proxy model for describing a probability mapping relationship between a target function and a parameter variable; determining a sampling function; and iteratively optimizing the to-be-optimized parameter variable by using the Bayesian statistical model and the sampling function to obtain an optimized parameter. The application can improve the coupling coordination degree of the complex system and meet the requirements of stable operation and intelligent control of the complex system.

The invention discloses an evapotranspiration Bayesian model average fusion method based on a hybrid statistical framework in China, and relates to the technical field of crossing of ecological hydrological remote sensing and Bayesian statistical modeling. According to the method, the four types of ET products are selected to be fused; the core algorithm can avoid structural deviation caused by a single model; the space-time coverage is complete, the space-time coverage covers the research area in 2007-2021, and the data consistency requirement of the fusion model is met; the reliability of a public data set which is widely used at home and abroad is widely verified in multiple independent researches. And the fusion process is realized through weighted average of the prediction probability density function of each model, and finally monthly ET estimation is generated.

This invention discloses a method and system for predicting the remaining lifespan of individual sensors based on Bayesian statistics, relating to the field of equipment health status monitoring and lifespan prediction technology. The invention provides a method comprising: establishing a general degradation model and prior distribution using historical degradation data; collecting monitoring data of a specific target sensor under real or accelerated stress in the field; updating the posterior distribution of model parameters based on Bayesian statistical inference and Gibbs sampling to generate a specific degradation model for that individual sensor; and calculating the remaining lifespan based on this specific model and a failure threshold. This invention achieves a breakthrough from "group lifespan assessment" to "accurate prediction of individual lifespan" through data-driven adaptive correction, significantly reducing over-maintenance costs and enhancing equipment operational safety.

The application provides a pathogen identification method based on a Bayesian model and application thereof, and the method comprises the following steps: collecting DNA information of a pathogen, screening SNP sites with higher consistency from the DNA information, and dividing the SNP sites into different cgSNPs combination types based on different pathogen typing; calculating the proportion of the reference genome of a certain pathogen typing in the reference genomes of all typing; calculating the proportion distribution of various pathogen typing in various cgSNPs combination types; and calculating the occurrence probability of the detected pathogen typing under the condition that a certain cgSNPs combination type is detected in the sequencing data analysis result through Bayesian statistics.

The invention provides a pipeline fatigue crack propagation behavior prediction method and system based on acoustic emission sensing. The method comprises the following steps: constructing a prediction initial model based on an obtained modeling parameter # imgabs0 # and a fatigue crack propagation rate # imgabs1 #; optimizing model parameters of the prediction initial model by adopting a Bayesian statistical method and a Markov chain Monte Carlo method to obtain optimal model parameters, and adjusting the prediction initial model based on the optimal model parameters to obtain a prediction model; and analyzing and predicting the acquired sample data based on the prediction model to obtain a prediction result. According to the method, the initial prediction model is constructed by integrating the modeling parameters and the fatigue crack growth rate, the initial model is optimized by using the Bayesian statistical method and the Markov chain Monte Carlo method, the prediction model is obtained, the prediction model can effectively predict behavior trends of cracks in different stages, and the prediction accuracy is improved. Potential safety hazards can be found in advance, and the overall safety and reliability of a pipeline system are remarkably improved.

The invention discloses a residential district spring water direct drinking pipe network hydraulic model construction and optimization design method. According to the method, a pipeline roughness coefficient slow-varying trend component is predicted based on a Langelier saturation index and a scaling rate physical equation, and a fast-varying disturbance component is extracted through variational mode decomposition, so that parameter multi-scale decomposition is realized; a physical information neural network is used as a hydraulic positive problem agent solver, whole-network hydraulic state parallel reasoning is completed within 50 milliseconds, and meanwhile an accurate analysis Jacobian matrix is obtained at zero extra cost through automatic differential; bayesian statistical fusion of multi-sensor observation and parameter estimation is realized through sparse ensemble Kalman filtering based on a pipe network observability coefficient, and pipeline roughness coefficient posterior probability distribution is output. According to the invention, parameter real-time calibration completely synchronous with the acquisition period of the sensor is realized, and the consumption of computing resources is reduced by more than 80%; water quality-hydraulic power double safety constraints are embedded; and whole-network high-precision state estimation under 20% of sensor coverage rate is supported.

The present invention discloses an event prediction and early warning method based on Bayesian deep learning, which mainly includes two parts: one is the construction of a Bayesian deep learning rule model based on event characteristics, and the other is the design of a prediction and early warning model based on the construction of the Bayesian deep learning rule model. By integrating Bayesian statistical methods with deep learning models, the present invention can fully utilize the advantages of both. It can combine prior knowledge and observation data to express the uncertainty of prediction results in the form of probability distribution, efficiently model and predict complex and nonlinear time series data, and achieve early warning of potential events, so as to improve the accuracy and reliability of event prediction and early warning.

The invention belongs to the technical field of product quality control, and relates to a product quality control method and system based on an optimization model. The method comprises the following steps: acquiring a nominal value, a confidence level and an error range of a defective rate of a product, determining an acceptance standard based on the nominal value, the confidence level and the error range, and calculating by using a binomial distribution formula to obtain a minimum detection sample size; based on the nominal value, the confidence level, the error range, the acceptance criterion and the minimum detection sample size, using Beta distribution to describe the prior distribution of the product defective rate; obtaining observation data after sampling detection based on the minimum detection sample size; based on the prior distribution and the observation data, the posterior distribution of the product defective rate is updated through a Bayesian formula; calculating a posterior probability that the product defective rate exceeds or is lower than the nominal value through a cumulative distribution function based on the nominal value, the prior distribution, the product defective rate and the posterior distribution; and establishing a Bayesian statistical model based on the prior distribution, the posterior probability and the observation data and solving the Bayesian statistical model.

The invention relates to the technical field of environment management, and discloses a carbon footprint evaluation method and device, and the method comprises the steps: carrying out the modeling of the probability distribution corresponding to the preprocessed activity data and background data based on a Bayesian statistical model, carrying out the sampling from the posterior distribution of the Bayesian statistical model through a Markov chain Monte Carlo method, and carrying out the sampling of the probability distribution of the Bayesian statistical model, probability distribution corresponding to the activity data is generated, carbon footprint calculation is carried out through target Monte Carlo simulation, random sampling is carried out in the probability distribution, an input variable combination is generated, distribution of carbon footprint calculation results is analyzed, and uncertainty analysis results are obtained; calculating the first-order sensitivity index and the total effect sensitivity index corresponding to the input variables, analyzing key variables in the carbon footprint evaluation result, and generating a global sensitivity analysis result; and constructing a carbon footprint evaluation model based on the uncertainty analysis result and the global sensitivity analysis result, and generating a carbon footprint evaluation result based on the carbon footprint evaluation model.

The present invention relates to the field of link transmission optimization technology, specifically to an intersatellite multipoint transmission and link dynamic optimization method based on Bayesian situational awareness. The method comprises the following steps: accessing an interstellar communication network interface, collecting real-time link data, including power, frequency, and signal-to-noise ratio, while recording the data transmission rate and delay of each link, as well as the packet loss rate, to generate real-time link status information. In the present invention, by monitoring the link status in real time and using Bayesian statistics for performance prediction, potential delays and anomalies are effectively predicted. By applying the Dijkstra algorithm to optimize the transmission path, the system adapts to the ever-changing environment of interstellar communication and ensures data transmission efficiency under extreme conditions. Dynamic routing adjustment strategies are updated according to real-time network load to ensure maximum data transmission efficiency. Automated network topology reconstruction further enhances the system's self-healing capability, reducing system downtime and potential performance degradation through rapid response.

The invention provides a pathogen identification method based on a Bayesian model and application of the pathogen identification method. The method comprises the steps that DNA information of pathogens is collected, SNP loci with high consistency are screened from the DNA information, and the SNP loci are divided into different cgSNPs combinations based on different pathogen types; calculating the proportion of the reference genome of a certain pathogen typing in all the typing reference genomes; calculating the proportion distribution condition of each pathogen subtype in each cgSNPs combination type; through Bayesian statistics, under the condition that a certain cgSNPs combination type is detected in a sequencing data analysis result, the occurrence probability of detected pathogen typing is calculated.

The present invention provides a fault detection method and system based on self-attention and neural differential equations, which belongs to the technical field of data analysis and machine learning. It solves the technical problem of efficiently identifying and detecting minor faults in dynamic and complex systems. Its technical solution is: including the following steps: 1) using the interquartile range IQR method to pre-process the collected sequence data to identify and eliminate outliers; 2) capturing long-distance dependencies in the sequence data; 3) constructing a neural ordinary differential equation model to analyze the extracted features; 4) performing a parameter search process based on a Bayesian statistical model to seek the optimal parameter combination. The beneficial effects of the present invention are: the present invention provides a novel, accurate and efficient detection technology solution for minor faults in dynamic and complex systems.

The present application provides a method for assessing aquatic ecological risks, a computer device, and a readable storage medium, which relate to the technical field of environmental governance. After obtaining the current aquatic ecological data, socioeconomic data, and historical aquatic ecological data of the target basin, the present application comprehensively considers the aquatic ecological risk factors that may be involved in the target basin by introducing the principles of Bayesian statistics and combining with the DPSIR model framework, and constructs an aquatic ecological risk assessment system based on the above-mentioned obtained data, so that the constructed aquatic ecological risk assessment system can more accurately express the internal relationships and interactions of the aquatic ecosystem at the target basin, to ensure the accuracy, credibility, and timeliness of the final aquatic ecological risk assessment results, facilitate researchers to more comprehensively and accurately understand the health status of the aquatic ecosystem at the target basin, and timely propose or adjust the governance decisions for the target basin.

The invention discloses a ground temperature field construction method based on Bayesian framework multi-source data fusion and integrated Kalman inversion, and relates to the technical field of ground temperature field calculation and geothermal parameter inversion, and the technical scheme is characterized in that the method comprises the following steps: obtaining preprocessed gravity and magnetic data; carrying out the inversion of the preprocessed gravity and magnetic data through the multi-source inversion of a Parker-Oldenburg algorithm and a LithoRef18 model, and carrying out the inversion of the preprocessed gravity and magnetic data; an additive Gaussian error model is established in combination with a heat conduction equation, posteriori distribution is taken as a core target under Bayesian statistics, an EKI algorithm is utilized to iteratively optimize parameters, and inversion parameters are output; and inputting the inversion parameters into a nonlinear heat conduction equation to obtain three-dimensional ground temperature fields with different depths, and reckoning surface heat flow distribution. According to the method, high-precision construction of the regional ground temperature field is achieved, compared with traditional interpolation and formula estimation, system errors are reduced, and the reliability of deep temperature and ground surface heat flow prediction is improved to a certain degree.

The application relates to the technical field of environmental management, and discloses a carbon footprint evaluation method and device, which comprises the following steps: modeling the probability distribution corresponding to the preprocessed activity data and background data based on a Bayesian statistical model; sampling from the posterior distribution of the Bayesian statistical model by using a Markov chain Monte Carlo method to generate the probability distribution corresponding to the activity data; generating an input variable combination by randomly sampling in the probability distribution through target Monte Carlo simulation to perform carbon footprint calculation; obtaining an uncertainty analysis result by analyzing the distribution of the carbon footprint calculation result; analyzing key variables in the carbon footprint evaluation result by calculating the first-order sensitivity index and the total effect sensitivity index corresponding to the input variables to generate a global sensitivity analysis result; and constructing a carbon footprint evaluation model based on the uncertainty analysis result and the global sensitivity analysis result, and generating a carbon footprint evaluation result based on the carbon footprint evaluation model.

This invention belongs to the field of computational materials and proposes a melting point calculation method based on first-principles molecular dynamics and Bayesian statistics. It employs a small-scale system for calculation, significantly reducing computational complexity compared to large-scale systems. Furthermore, it utilizes an on-the-fly machine learning force field to further accelerate melting point calculation. This method requires no manual intervention, automatically updating the force field during AIMD simulations to maintain the accuracy of first-principles calculations and significantly improve computational efficiency. This method calculated 16 solid-liquid configurations within the 800–1100 K range, consuming approximately 600 cores, which significantly shortens computation time compared to other methods, thus demonstrating great application potential.

The invention provides a rainfall-type landslide susceptibility evaluation method, system and device and a medium, and belongs to the technical field of geological disaster risk evaluation and spatial information modeling. The method comprises the following steps: collecting and preprocessing multi-source environment and rainfall data of a target area; constructing a multi-time scale rainfall factor system, and reducing variable colinearity through intra-group screening and a multi-stage feature engineering method; modeling the static environment factor by using a random forest model, obtaining a landslide background susceptibility probability, and converting the landslide background susceptibility probability into a logarithmic probability offset item; under the constraint of an offset term, constructing a Bayesian spatial statistical model, introducing a nonlinear effect term and a spatial random effect term of a rainfall factor, and carrying out joint modeling and Bayesian inference on the occurrence probability of the landslide; and landslide susceptibility dynamic prediction and spatial mapping are realized. According to the method, the nonlinear prediction capability of machine learning and the physical constraint and spatial modeling capability of Bayesian statistics are fused, so that the stability, prediction precision and mechanism interpretation of an evaluation result are improved.

Embodiments of the present invention disclose a method, apparatus, device, and medium for estimating the plaintext similarity of encrypted strings. The method includes: obtaining a plaintext data set, performing an encryption operation on the plaintext data set using a preset encryption algorithm to obtain a ciphertext data set; respectively modeling the plaintext data set and the ciphertext data set based on a multinomial distribution to obtain an estimated distribution corresponding to the plaintext data set and an estimated distribution corresponding to the ciphertext data set; based on a Bayesian statistical model, estimating the estimated distribution corresponding to the decryption function according to the estimated distribution corresponding to the plaintext data set and the estimated distribution corresponding to the ciphertext data set; estimating the plaintext similarity between different target encrypted strings according to the estimated distribution corresponding to the decryption function. By adopting the above technical solution, the similarity of the plaintext data before encryption can be estimated through the encrypted ciphertext data, and while protecting the privacy of the plaintext data, the correlation relationship between multiple plaintext data can also be estimated.

The application discloses a kind of power transmission line defect monitoring method and system based on multi-source data fusion, it is related to electric power equipment state monitoring field, solved the technical problem that defect detection method reliability is insufficient, false alarm rate is high, lack of accurate positioning to defect reason.The application collects the multi-source sensor data of power transmission line, based on the measured physical characteristic value of measured component of multi-source sensor data extraction, and based on environmental parameter and electrical parameter calculation theoretical physical characteristic value, form feature pair set;Each feature pair is judged based on bayesian statistical test method to abnormal state, and the abnormal type identification is output;Abnormal type identification and multi-source sensor data are used as observation data, input into the defect causal diagram model of preposition and carry out probability reasoning, and defect type is determined according to maximum posterior probability principle;Based on defect type, call physical evolution model to predict the future evolution trend of defect state, and assess risk level and the remaining useful life of measured component.

The invention relates to a battery car anti-collision early warning method and system based on an intelligent algorithm. The method comprises the steps that a dynamic image and a static image of an obstacle around a battery car and the distance between the battery car and the obstacle are obtained through image obtaining equipment and an ultrasonic sensor; and taking the environmental data as the input of the FusionNet multilayer perceptron model to call the FusionNet multilayer perceptron model to learn and analyze the environmental data so as to identify the obstacle and predict the moving path of the obstacle. And calling a dynamic risk assessment algorithm based on Bayesian statistics to calculate a risk assessment result of collision between the battery car and the obstacle according to the moving path of the obstacle, and adjusting a collision early warning level and obtaining an anti-collision measure according to the risk assessment result. And when the collision early warning level exceeds the preset level, an early warning signal is sent based on the anti-collision measures, and a reaction mechanism of the battery car is started, so that the battery car is prevented from colliding with the obstacle.