35 results about "Generalized linear model" patented technology

The invention provides a tea garden drought prediction method and system fusing LightGBM and LSVM in the technical field of smart agriculture. The method comprises the steps of S1, collecting a large amount of historical monitoring data to construct a data set; s2, the data set is analyzed through a Limma algorithm, a COX survival regression model, a GLM generalized linear model and an SHAP, and an environment factor correlation graph is obtained; s3, creating a tea garden drought prediction model; s4, dividing the data set into a training set, a verification set and a test set to train, verify and test the tea garden drought prediction model, and deploying the tea garden drought prediction model passing the test; and S5, collecting real-time monitoring data from the tea garden, inputting the data into the tea garden drought prediction model to obtain a tea garden drought prediction result, and displaying the tea garden drought prediction result on a visual interface. The tea garden drought prediction method has the advantages that the timeliness, accuracy and generalization ability of tea garden drought prediction are greatly improved.

The invention relates to a method and a device for mapping a cortical nerve image marker in a neurotransmitter map, which can be used for exploring abnormal modes of cortical structures and functions of SCA patients, carrying out functional annotation on the anomalies and analyzing the association between the anomalies and a specific biological process. The method comprises the following steps: (1) obtaining 3D-FSPGR and GRE-SS-EPI sequences; (2) preprocessing image data; (3) image feature extraction: dividing a cortex region through a network function partition map; in order to evaluate the difference between the groups, a t graph of each feature is calculated based on a generalized linear model (GLM), and the t graph comprises the cortex thickness, the brain sulcus depth, the curvature, the ALFF and the ReHo; the model takes gender, age and total cranial volume as covariables so as to control potential confounding factors; and (4) associating the Neuromaps annotation with the structural and functional characteristics.

Systems, methods, and computer-readable media are provided for determining matches between records of different systems based on aggregate record data, and graphically marking potentially matched groups of data along with predicted confidence levels. Preliminary matching tools may allow allow users to define various rules based on which a majority of the transactions can be matched and reconciled. However, remaining transactions are disposed of in an interactive matching process. The matches may be determined unidirectionally from a source transaction to transactions from a target ledger, or bidirectionally from transactions in the target ledger to transactions other than the source transaction. Transactions may be matched many-to-many, one-to-many, or many-to-one, and a proposed order of match selections may be presented in a user interface. Match metadata or insights may be displayed to show a confidence of the match, reasons for the confidence, and / or a confidence of other matches that may be more beneficial than a match with a source transaction. The confidence and match insights may be generated by a machine learning model with access to transactions from a source transaction ledger and a target transaction ledger. The machine learning model may be trained on manual activity for prior matches that have been made. Matches may be performed using a hybrid machine learning model that accounts for random forests, decision trees, neural networks, naïve bayes algorithm, and / or a generalized linear model. Machine learning models also incorporate ongoing feedback from the users who can either accept or reject suggested matches and hence the models undergo an evolution process and constantly update from user patterns.

The invention discloses a credit scoring model modeling method based on transfer learning, and the method comprises the steps: a data collection stage: dividing a sample into a source domain and a target domain, and dividing the sample of the target domain into a training set and a test set; in the data preprocessing stage, special values and abnormal values are removed, initial weights are given to samples, linearization based on the evidence weight (WOE) is carried out on variables, and then variables suitable for model entering are screened according to the distinguishing capacity of the variables for good and bad samples and correlation between the variables; in the model iteration stage, based on the improved migration adaptive enhancement algorithm (TrAdaBoost.R2) aiming at the regression problem as a basic framework, a generalized linear model is adopted as a reference model, and repeated iteration is carried out on the weight and the model; in the iteration stopping stage, whether iteration is stopped or not is judged based on the model loss function and the expression trend of the model loss function on the test set. According to the method, credit scoring model modeling based on small samples is realized, and the speed of model iteration is greatly improved.

This invention relates to apparatus and methods for selecting a soil microbiome having increasing plant productivity. The apparatus and methods relate to combining an output from a machine learning model, a generalized linear model, and a distance-based multivariate model to execute a donor forest selection tool that is configured to select from a plurality of geographically distinct plant communities a soil microbiome having increasing plant productivity.

The invention provides a tea garden drought grading prediction method and system based on machine learning in the technical field of smart agriculture. The method comprises the steps of S1, collecting a large amount of historical monitoring data from a tea garden; s2, preprocessing each historical monitoring data and then constructing a data set; s3, analyzing the data set through an L-imma algorithm, a COX survival regression model and a GLM generalized linear model to obtain an environmental factor correlation graph; s4, creating a tea garden drought prediction model based on the multi-modal feature extraction layer, the feature fusion layer, the soil entropy prediction layer and the drought mapping output layer; s5, training the tea garden drought prediction model through the data set, and deploying the trained tea garden drought prediction model; and S6, predicting the drought of the tea garden through the deployed tea garden drought prediction model. The method has the advantages that the accuracy of tea garden drought prediction and the refinement degree of irrigation decision are greatly improved.

The invention provides a time-space geographically weighted regression method and device for counting data, equipment and a storage medium, and relates to the technical field of time-space information data analysis. The method comprises the following steps: acquiring multi-source space-time observation data of a target area; constructing a geographically weighted Poisson regression model or a geographically weighted logistic regression model fused with space-time non-stationarity based on a generalized linear model framework so as to fit a local space-time variation relationship between a numeric type or binary type response variable and an independent variable; solving a regression coefficient of the spatio-temporal change through an iterative algorithm; and finally, model output is converted into early warning information, a dredging scheme or a planning decision. According to the method, the data discrete characteristics and the process spatial-temporal heterogeneity are described in a unified manner, so that the problems of modeling error and insufficient precision caused by neglecting the spatial-temporal coupling effect and distribution mismatch when a traditional model is used for coping with spatial-temporal data with discrete distribution characteristics are solved; the accuracy and decision support capability of dynamic risk early warning and refined resource allocation in the fields of environmental monitoring, traffic safety, land planning and the like are remarkably improved.

The invention provides a method and a system for predicting a potential suitable growth area of alien invasive species in the technical field of risk management of invasive species. The method comprises the following steps: S1, acquiring a large amount of environment variable data and species distribution sample data; s2, downscaling the environment variable data through a spatial regression model to construct a high-resolution environment data set, performing spatial refinement on the species distribution sample data through a Moran index to construct a sample data subset, and constructing a data set based on the high-resolution environment data set and the sample data subset; s3, creating a species distribution model based on a maximum entropy model, a random forest, a support vector machine, a generalized linear model, an XGBOOST or a deep learning model, and training the species distribution model through the data set; and S4, predicting the suitable growth area of the alien invasive species through the trained species distribution model. The method has the advantages that the prediction precision and generalization ability of the potential suitable growth area of the alien invasive species are greatly improved.

The present disclosure relates to an application of gene markers in multi-cancer early detection, a method for constructing an early detection model, and a detection device. In the present disclosure, low-coverage whole-genome sequencing is conducted on cell-free DNAs (cfDNAs) from a plasma sample, and according to high-throughput sequencing results, six differential features of the cfDNA fragments are analyzed for each cancer. Then the training and modeling are conducted with a convolutional neural network to allow the early detection of a plurality of cancers at a low sequencing depth. Then the training and modeling are conducted with a generalized linear model (GLM), a gradient boosting machine, a random forest model, a deep learning model, and an extreme gradient boosting model, and staking is conducted with a GLM to construct a multi-feature algorithm, to allow the tissue-of-origin-based detection of cancers.

The invention discloses a machine learning-based tumor early screening expiration VOCs marker identification and verification method. The method comprises the steps of obtaining and processing high-dimensional full-spectrum expiration VOCs data; a generalized linear model is adopted to evaluate the influence of the queue correlation difference on the expiratory VOCs data, and VOCs with significant interaction are removed; screening potential expired VOCs biomarkers with significant difference between an experimental group and a control group by adopting a difference volcano plot method in which a kernel weighting function is introduced; a Boruta feature screening algorithm based on a random forest is adopted, and an expiration VOCs biomarker panel most related to the tumor is constructed; and evaluating the tumor diagnosis and grading performance of the expired VOCs biomarker panel by adopting a machine learning algorithm based on multiple underlying logics. According to the method, a marker panel construction process based on a machine learning algorithm is created, and a biomarker panel suitable for various algorithms is accurately and efficiently captured from massive expired VOCs.

The application discloses a method for predicting eel fry resource quantity, comprising the following steps: obtaining initial factors, screening the initial factors to obtain influence factors, obtaining historical data according to the influence factors, and constructing a generalized linear model and a nonlinear model; optimizing and verifying the generalized linear model and the nonlinear model through the historical data, obtaining current data according to the influence factors, predicting the current data through the verified generalized linear model and the nonlinear model, and combining the prediction results by weighting to generate an eel fry resource quantity prediction result. Through the technical scheme, the eel fry resource quantity in an estuary can be effectively and universally predicted.

The invention provides a visual monitoring method and system for tea garden drought in the technical field of intelligent agriculture. The method comprises the steps that S1, a large amount of historical monitoring data is collected from a tea garden through a sensor array; s2, preprocessing each historical monitoring data and then constructing a data set; s3, analyzing the data set through an L-imma algorithm, a COX survival regression model and a GLM generalized linear model to obtain an environmental factor correlation graph; s4, creating a tea garden drought prediction model based on the environmental factor correlation graph; s5, training and deploying the tea garden drought prediction model through the data set; s6, collecting real-time monitoring data from the tea garden, and inputting the data into the tea garden drought prediction model to obtain a tea garden drought prediction result; and S7, converting the tea garden drought prediction result into a drought thermodynamic diagram for display. The method has the advantages that the accuracy of tea garden drought prediction, the refinement degree of irrigation decision and the intuition of drought display are greatly improved.

Systems, methods, and computer-readable media are provided for determining matches between records of different systems based on aggregate record data, and graphically marking potentially matched groups of data along with predicted confidence levels. Preliminary matching tools may allow allow users to define various rules based on which a majority of the transactions can be matched and reconciled. However, remaining transactions are disposed of in an interactive matching process. The matches may be determined unidirectionally from a source transaction to transactions from a target ledger, or bidirectionally from transactions in the target ledger to transactions other than the source transaction. Transactions may be matched many-to-many, one-to-many, or many-to-one, and a proposed order of match selections may be presented in a user interface. Match metadata or insights may be displayed to show a confidence of the match, reasons for the confidence, and / or a confidence of other matches that may be more beneficial than a match with a source transaction. The confidence and match insights may be generated by a machine learning model with access to transactions from a source transaction ledger and a target transaction ledger. The machine learning model may be trained on manual activity for prior matches that have been made. Matches may be performed using a hybrid machine learning model that accounts for random forests, decision trees, neural networks, naïve bayes algorithm, and / or a generalized linear model. Machine learning models also incorporate ongoing feedback from the users who can either accept or reject suggested matches and hence the models undergo an evolution process and constantly update from user patterns.

The present application belongs to the technical field of cosmetic safety evaluation, and discloses a human patch test result typing method for cosmetics suitable for sensitive skin, which comprises the following steps: collecting comment data and performing deduplication and noise reduction processing, analyzing the comment data set using a trained RoBERTa-large language model, identifying adverse reaction comments, and calculating adverse reaction scores and adverse reaction rates; performing a patch test and setting a typing line according to the patch test results; extracting product characteristics of the cosmetics; establishing a generalized linear model; testing the likelihood ratio chi-square and significance of each generalized linear model, and screening the best typing of the cosmetics corresponding to the patch test. The method of the present application can effectively evaluate the safety risk of cosmetics for target users.

The application discloses a monochrome sensor mixed coding illumination and image reconstruction method and an imaging system, relates to the technical field of computational imaging, and comprises the following steps: first, at least three RGB mixed illumination modes are designed, and a coding matrix is constructed; then, a plurality of frames of gray scale images under the same target state are collected through a global shutter monochrome sensor according to sequence exposure; subsequently, a generalized linear model is constructed based on the coding matrix, an equation set is solved, and a color image is reconstructed; finally, a standard format image is output after post-processing. The application can eliminate dynamic color smearing in principle, has frame-level self-calibration capability, is compatible in algorithm and strong in robustness, keeps the system small and low in cost, and is suitable for dynamic imaging high requirement scenes such as medical and industrial scenes.

The invention discloses a hospital quality inspection management system and method based on AI intelligent analysis, and the method comprises the steps: collecting quality inspection data, and carrying out the preprocessing of the quality inspection data to generate a standardized data set; carrying out feature coding and format conversion, and constructing a quality inspection feature matrix with a unified format; sparse feature selection, feature importance evaluation and key feature screening; constructing an improved generalized linear model, and mapping key features and a target index function; performing parameter fitting and tuning to obtain a generalized linear model with optimal performance; performing generalized linear model prediction, and outputting a comprehensive quality analysis result; and generating a structured analysis report, intelligent early warning, task assignment and closed-loop management. According to the invention, by introducing the improved generalized linear model and the feature optimization mechanism, intelligent analysis and closed-loop management of hospital quality inspection data are realized, and the accuracy of quality early warning and the rectification response efficiency are improved.

The invention relates to the technical field of spine assessment and treatment, in particular to a children spine form assessment and detection method based on multi-source data fusion. Comprising the following steps: S1, obtaining observation indexes of a to-be-detected child and performing preprocessing; s2, constructing a potential section analysis model based on the standardized value, obtaining a fitting index of each model, and carrying out weighted discrimination on the fitting indexes to determine an optimal potential category number; and S3, performing variance analysis on each potential category among the standardized values of the observation indexes to check differences, and summarizing phenotypic features of each potential category according to difference directions and effect sizes of the indexes among different potential categories and naming the phenotypic features. According to the method, the adjusted generalized linear model is adopted on single pollutant analysis, and weighted quantiles, regression and multi-interpolation expansion thereof are introduced on mixed exposure analysis, so that the method is more targeted in risk identification and intervention target selection.

The application discloses a kind of method and system of joint parameter estimation and signal reconstruction under distributed antenna system, method includes: S1: constructing the generalized linear model of parameter estimation and signal reconstruction under distributed antenna system;S2: noise variance parameter and signal estimation value initialization, set iteration stop condition and obtain the input value required by preset algorithm;S3: respectively obtain the input parameter required by the noise variance estimator of each cluster antenna end and the approximate posterior probability of signal;And the estimated value of signal is solved;S4: the noise variance parameter corresponding to signal is estimated independently by minimizing variational bet free energy;S5: whether the preset iteration stop condition is reached, if yes, then flow to step S6;If not, then return to step S3 and carry out next round iteration;S6: iteration ends, and the estimated value of signal and the estimated value of corresponding noise variance of each cluster antenna end are output.The application realizes the joint estimation of noise variance and signal under different environments, and improves signal reconstruction performance.

The invention discloses a data-driven demand response feature recognition and uncertainty quantification method in a power system, and the method comprises the steps: firstly collecting the historical demand response data of a user, and constructing a data set comprising a user feature vector X, a response intention label Y and a response potential label Z; and establishing a mapping relation between the user feature vector X and the response will probability by using a generalized linear model, and assuming that Y obeys Bernoulli distribution. Aiming at a user with a response intention, establishing a mapping relation between conditional expectation of a response potential label Z and X through a generalized linear model framework, and constructing a parameterized probability distribution model of the Z; on the basis, joint probability distribution of Y and Z is constructed. And based on a sample independent identically distributed hypothesis, constructing a logarithm joint likelihood function, obtaining optimal parameter estimation by adopting a gradient descent algorithm, and determining an optimal model through AIC and BIC criteria. And finally, quantifying the uncertainty of response willingness and potential, and constructing a comprehensive index to evaluate the reliability degree of user response.

The invention discloses a cascade hydropower station surrounding environment effect evaluation integration method and a related device, and belongs to the technical field of environmental science and engineering. The method comprises the following steps: acquiring NDVI data and related climate data before and after a cascade hydropower station is built, and calculating to obtain annual vegetation coverage time sequence raster data and related annual climate time sequence raster data; converting the time sequence raster data into numerical data in a text form; calculating according to the numerical data to obtain a partial correlation coefficient of the vegetation coverage on each related climate factor, performing T test, and analyzing the influence of each related climate factor on the vegetation coverage; constructing a generalized linear model, obtaining a reservoir construction state, inputting the numerical data and the reservoir construction state into the generalized linear model for fitting, and evaluating contribution conditions of each factor to vegetation; and constructing a path model, inputting the numerical data and the reservoir construction state into the path model, and quantifying the influence degree of each factor on the NDVI.

The present invention discloses an analysis method, system and storage medium capable of controlling kinship correlation in a sample, comprising the following steps: S1: acquiring sample data; S2: constructing a generalized linear model based on the sample data, estimating the parameters of a fitted constraint model and calculating residuals; S3: dividing the residuals of the fitted constraint model into outliers and non-outliers, calculating and storing the homology sharing probability between any two members in a family where the number of members is greater than 1 and the sample data corresponding to the outlier is located; S4: dividing all sites in the whole genome into a certain number of intervals according to the minor allele frequency, calculating and storing the joint probability of the genotype vectors of each family member at the endpoints of each interval; S5: calculating the statistical p-value for each site to be tested using a hybrid test strategy combining a normal distribution approximation method with an empirical saddle point approximation method; the analysis method avoids a large number of false positive or false negative results.

The present invention provides a method for monitoring column performance and operating a chromatography column by applying a generalized linear model to system suitability parameters (SSPs) to assess how quickly the column is changing over time and whether the column stationary phase needs to be replaced. The method leads to faster identification of column failures, helping to maintain high separation quality and consistent analytical results for analytical and preparative chromatography methods. Columns evaluated and / or monitored by the method, as well as products resulting from the use of these columns and methods, are also provided.

An apparatus for generating a generalized linear model structure definition by generating a gradient boosted tree model and separating each decision tree into a plurality of indicator variables upon which a dependent variable of the generalized linear model depends.

The invention provides a respiratory tract epidemiological social contact gathering point identification and action range calculation method, and relates to the technical field of public health and epidemiology, and the method comprises the steps: obtaining infection case data and city POI data, carrying out the preprocessing of the infection case data, carrying out the spatial pairing of the preprocessed infection case data and a research region, carrying out the arrangement of the POI data, and selecting a social contact gathering point; a basic space unit is determined, the number of infected cases and the social contact gathering point density are counted and preprocessed, and multi-collinearity testing is carried out; establishing a generalized linear model of Poisson connection, and optimizing the linear model in combination with stepwise regression to obtain key social aggregation points; and constructing a multi-scale geographic weighting model, and calculating the optimal bandwidth of each key social aggregation point. According to the method, the key social contact gathering points influencing the transmission of respiratory tract infectious diseases can be accurately identified, the spatial action range of the key social contact gathering points is quantified, and a scientific basis is provided for formulating differentiated prevention and control strategies.

The invention belongs to the technical field of cosmetic safety evaluation, and discloses a cosmetic human body patch test result typing method suitable for sensitive skin, which comprises the following steps: collecting comment data and carrying out de-weighting and noise reduction processing, analyzing a comment data set by using a trained RoBERTa-large language model, identifying adverse reaction comments, and classifying the adverse reaction comments. Calculating an adverse reaction score and an adverse reaction rate; performing a patch test, and setting a parting line according to a patch test result; extracting product characteristics of the cosmetics; establishing a generalized linear model; and checking the likelihood bichi square and significance of each generalized linear model, and screening the optimal classification of the cosmetics corresponding to the patch test. According to the method, the safety risk of cosmetics to the target user can be effectively evaluated.

The present invention discloses a signal reconstruction method based on the empirical Bayesian method for processing unknown transition probabilities or unknown signal priors and transition probabilities. For the problem of unknown transition probabilities in generalized linear models, a transition probability model based on empirical Bayesian estimation is established to obtain the transition probability of the generalized linear model. According to the estimated transition probability and the state update equation, a scalar model is obtained, and a probability model based on empirical Bayesian estimation priors can be established to obtain the prior distribution of unknown signals. The transition probability model based on empirical Bayesian estimation and the probability model based on empirical Bayesian estimation priors are added to the algorithm to perform signal recovery. The algorithm continuously updates the signal estimation value through iterative calculation until the convergence condition is reached, thereby realizing accurate recovery of the original signal. The present invention solves the problem of unknown transition probabilities or unknown signal priors and transition probabilities through empirical Bayesian estimation, and can also reconstruct signals.

The invention discloses a false data injection attack real-time protection method for a dynamic line rating system, which belongs to the technical field of intelligent power grid network security, and comprises the following steps of: 1, collecting environmental data and preprocessing the data; the method comprises the steps of (1) obtaining a historical data matrix, (2) detecting daily environment data through a z-score method to obtain daily data, and updating the historical data matrix according to the daily data, (4) constructing a training data set and a test data set, (5) constructing a z-score-BGLM-LR model through a machine learning classification algorithm of binary generalized linear model logistic regression, (6) training the model by using the training data set, and (7) obtaining a z-score-BGLM-LR model. And 7, training the model, testing the model by using the test data set after the evaluation is completed, and detecting the real-time protection capability of the model, the detection model is constructed by adopting the method, and the method is suitable for network security protection in the dynamic line rating operation environment of the power system.

The application discloses a pet intelligent delivery room, which comprises a medical box for accommodating pets, and the medical box is provided with a database, a collection module, a processing module and a warning module; the risk assessment process of the pet intelligent delivery room comprises the following steps: screening and classifying extraction of historical data of pet delivery processes previously imported into the database to determine risk related factors in the delivery process; the risk related factors are valued to obtain corresponding quantitative data, the quantitative data are preprocessed, and multiple ordered logistic regression analysis is performed on the processed data to obtain a logistic regression model; generalized linear model analysis is performed on the quantitative data, and the regression model is evaluated; physiological indexes of the pet delivery process are collected in real time, and a risk category is predicted; and when the risk assessment reaches a warning condition, an alarm information is sent through the warning module. The application can effectively predict and monitor the risk in the pet delivery process.

The present disclosure relates to an application of gene markers in multi-cancer early detection, a method for constructing an early detection model, and a detection device. In the present disclosure, low-coverage whole-genome sequencing is conducted on cell-free DNAs (cfDNAs) from a plasma sample, and according to high-throughput sequencing results, six differential features of the cfDNA fragments are analyzed for each cancer. Then the training and modeling are conducted with a convolutional neural network to allow the early detection of a plurality of cancers at a low sequencing depth. Then the training and modeling are conducted with a generalized linear model (GLM), a gradient boosting machine, a random forest model, a deep learning model, and an extreme gradient boosting model, and staking is conducted with a GLM to construct a multi-feature algorithm, to allow the tissue-of-origin-based detection of cancers.

The application provides a personalized book recommendation method based on a Wide&Deep model. Previous studies mainly rely on manual analysis of readers' reading history, book classification and feedback to provide recommendation services for users, but there are problems of insufficient personalization and low efficiency. In order to solve this problem, a recommendation system combining generalized linear models and deep neural networks is proposed. The Wide part helps the model capture common behavior patterns of users by processing handcrafted features, while the Deep part is used to process high-dimensional sparse features and capture complex relationships between user interests. After preprocessing the user data, the Wide&Deep model is input to generate personalized recommendation results.