29 results about "Receiver operating characteristic" patented technology

Systems, methods, and apparatuses for selecting a model are described. A method includes receiving a request to perform model selection and evaluating multiple models by generating various metrics for each trained model. These metrics include a forewarning time (how much in advance of a failure an alert can be raised), an event recall metric (how many failure events were alerted to in advance), an event precision metric (ratio of true and false positives, and an area under a receiver operating characteristic (ROC) curve. For each trained model, a weighted harmonic mean is calculated from these metrics. A model is then selected based on the calculated means and a report on the selected model is generated and provided.

Systems and methods for predicting esophageal adenocarcinoma (EAC) and gastric cardia adenocarcinoma (GCA) using machine learning are provided. An example system may obtain an electronic health record (EHR) dataset, identify missing values in the EHR dataset, and generate imputed values for the missing values using simple random sampling imputation. The system may train a model using an extreme gradient boosting algorithm and a training dataset including the EHR dataset to generate a trained model including multiple decision trees. Training the model includes tuning the model to achieve a greatest value of an area under a receiver operating characteristic curve associated with the model. The system may obtain a patient EHR dataset, generate a prediction associated with a risk of EAC and / or GCA by applying the trained model to the patient EHR dataset, and provide the prediction to a computing device to determine a patient treatment protocol.

The invention discloses a pulmonary nodule algorithm performance test method, device and equipment and a storage medium, which are applied to the technical field of artificial intelligence, and comprise the following steps: determining a test method of a target detection scene and a test method of a segmentation and measurement scene as test methods; determining at least one of a selection matching mode, a first recall rate, first accuracy, F1 measurement, average accuracy, an average accuracy mean value and a free response subject operation characteristic curve as a test value; determining and selecting at least one of a second recall rate, a second accuracy, a target area overlap ratio, tree length detection and a surface distance as a test value; and testing the pulmonary nodule algorithm to be detected according to the test data set, and determining a performance test result according to the test result and the test index threshold. The algorithm is tested according to the test method of the target detection scene and the test method of the segmentation and measurement scene, and the algorithm performance is determined according to the test result and the threshold thereof, so that the pulmonary nodule algorithm performance can be detected.

The application provides an intercept rendezvous condition analysis method and system based on a binary decision tree, comprising: establishing a rendezvous condition mathematical model; setting intervals of a rendezvous angle, an initial relative distance and a target speed ratio of a spacecraft in the rendezvous condition mathematical model according to random distribution characteristics of measurement noise, target inclination and radial maneuver noise of a spacecraft detection device, and generating training data through a relative motion model; training a binary decision tree model by using the labeled training data, evaluating the obtained binary decision tree model according to a receiver operating characteristic (ROC) curve, and obtaining a trained binary decision tree model when the evaluation meets preset requirements; and extracting rendezvous condition rules from the binary decision tree according to constraint conditions and bifurcation endpoint data of each layer node of the trained binary decision tree model, and obtaining set-form rendezvous condition rules with interpretability through set intersection and union operations of intervals of each element in the rendezvous condition.

The present invention discloses a method for predicting steel breakout based on stacking multi-classifier fusion, belonging to the field of iron and steel metallurgy. The method comprises the following steps: constructing a sample library of true and false steel breakouts; extracting temperature characteristic data and time-series temperature rates of true and false steel breakouts; constructing feature vectors of true and false steel breakout samples; preprocessing the true and false steel breakout feature data; constructing a steel breakout prediction model based on stacking multi-classifier fusion using random forest, K-nearest neighbor classification, and support vector classification as primary classifiers, and logistic regression as a secondary classifier; and finally determining an optimal threshold point through a receiver operating characteristic curve. If the threshold is greater than the optimal threshold, the prediction is judged as steel breakout. The present invention combines the steel breakout temperature characteristic vector with a multi-classifier fusion recognition method to establish a strong classification and recognition model for steel breakouts. While ensuring a 100% steel breakout reporting rate, the method reduces the false alarm rate and improves the prediction accuracy.

The invention belongs to the technical field of bioinformatics and medical detection, and particularly relates to a marker combination for neuroblastoma (NB) noninvasive risk level grading, a prediction model, a prediction method and application thereof. The marker combination is used for determining the sex, determining whether the month age is greater than 18 months, determining whether plasma MYCN is amplified, determining whether tumors are metastatic, and determining the content of neuron-specific enolase and lactic dehydrogenase; a machine learning algorithm is used for constructing an NB noninvasive risk degree grading prediction model, the comprehensive performance of the random forest model is optimal, the area value under a subject working characteristic curve reaches 0.956, the sensitivity is 92.9%, the specificity is 82.1%, the accuracy rate is 87.5%, the Kappa value is 0.75, the F1 score is 0.881, and NB middle and low risk patients and NB high risk patients can be effectively distinguished; the NB non-invasive risk level grading prediction model constructed by the invention can quickly, accurately and non-invasively perform NB risk level grading, and has a relatively good clinical application value.

The present invention relates to the field of biomedicine, and more specifically, to a prediction model and method for hypertrophic choroidal disease. The prediction model is constructed by the following steps: S1) Data acquisition: selecting eyes with hypertrophic choroidal disease, eyes with normal hypertrophic choroids, and eyes with normal non-hypertrophic choroids as data collection objects, and collecting the choroidal thickness in the macular region, the thickness of the choroidal medium vascular layer, and the thickness of the choroidal large vascular layer; S2) Data processing: calculating the ratio of the Sattler layer thickness to the Haller layer thickness to obtain the S / H ratio; comparing and analyzing the choroidal thickness and S / H ratio across age groups and groups, and establishing a multivariate regression model; S3) Data output: drawing a receiver operating characteristic (ROC) curve to obtain a prediction model for hypertrophic choroidal disease based on the S / H ratio. This objectively reflects the comparison with healthy eyes in the normal population and those with hypertrophic choroidal disease.

The invention provides a multi-angle persistent chemical screening method based on a multi-task graph neural network, and belongs to the technical field of high-throughput screening for chemical risk management. The method aims at achieving synchronous, efficient and accurate screening of the related attributes of the durability of the chemicals. The method comprises the following steps: firstly, constructing a multi-angle chemical durability parameter data set; the method comprises the following steps: taking a molecular graph generated by a chemical structure as model input, constructing a model through a graph neural network, and finishing synchronous training by adopting a mean square error and a superposition loss function with positive sample weight cross entropy; evaluating the performance of the model through indexes such as a determination coefficient and an area under a subject working curve; and characterizing a model application domain by adopting structure-activity relationship morphology analysis. The model constructed by the research relates to more persistent parameters, is more comprehensive, larger in data volume and wider in coverage range, has good fitting ability, prediction ability, robustness and reliability, and can provide technical support for environmental risk assessment of persistent chemicals.

The present invention discloses a coronary artery calcification prediction and analysis system based on the receiver operating characteristic curve, which relates to the field of coronary artery calcification detection technology, including: an integrated processing unit for collecting clinical data, laboratory data, and imaging data, and generating standardized data through data preprocessing; an evaluation and prediction unit for predicting the risk probability of coronary artery calcification under different physiological feature combinations and supporting risk stratification; an analysis and display unit for generating transparent explanations and assisting decision-making through interactive visualization tools; a decision optimization unit for generating intervention recommendations based on the prediction results and optimizing blood magnesium management strategies. The present invention significantly improves the prediction accuracy and clinical practicality through an improved random forest model that integrates dynamic integrated selection and weighted probability. Through algorithm innovation and visual closed-loop design, it achieves full-chain optimization from data to decision-making, providing a precise, transparent, and dynamic intelligent tool for the prevention and treatment of CAC in CKD patients.

This invention provides an intelligent fault prediction and diagnosis method for chemical processes, comprising the following steps: S1. Data acquisition: Collecting monitorable data from chemical processes using chemical sensors; S2. Data preprocessing: Decomposing the collected data into spatial scales using variational mode decomposition to obtain the positional dependencies of different sensors, and iteratively optimizing the decomposition parameters using a beaver optimizer; S3. Constructing a deep learning model for time-frequency domain feature fusion: Based on a common convolutional block attention module, an improved feature-injected convolutional block attention module is proposed; based on a temporal convolutional neural network, an improved frequency-domain convolutional module is proposed to further extract features; S4. Deep learning model training: Using a cross-entropy loss function and an Adam optimizer, the preprocessed data is divided into training, validation, and test sets and input into the model for multiple training iterations; S5. Evaluation of diagnostic results: The diagnostic performance of the model is comprehensively evaluated using recall, F1 score, and area under the receiver operating characteristic curve, and the best-performing model is selected from multiple trained models for deployment. This method enables real-time extraction and fusion of time-frequency domain features from chemical sensors, achieving efficient fault prediction and diagnosis in chemical processes.

Systems and methods for detection and localization of image and document forgery. The method can include the step of receiving a dataset having a plurality of authentic images and a plurality of manipulated images. The method can also include the step of benchmarking a plurality of image forgery algorithms using the dataset. The method can further include the step of generating a plurality of receiver operating characteristic (ROC) curves for each of the plurality of image forgery algorithms. The method also includes the step of calculating a plurality of area under curve metrics for each of the plurality of ROC curves. The method further includes the step of training a neural network for image forgery based on the plurality of area under curve metrics.

The invention discloses a method and a system for distinguishing benign and malignant ovarian lumps based on RegNet of a vaginal ultrasound image in combination with triple ROI constraints. According to the method, RegNetY-032 is adopted as a backbone network, an intelligent group level processing strategy and a four-level enhancement mechanism are combined, and it is ensured that a model focuses on ROI through triple constraints of input level ROI mask, feature level ROI constraint and ROI outer weight penalty loss. The device domain self-adaption technology based on the gradient inversion layer is further achieved, and the problem of imaging difference between different ultrasonic devices is effectively solved. According to the system, multi-angle image fusion is carried out by adopting a self-attention mechanism, the area of a subject under a working characteristic curve ROC in external verification reaches 0.992, the accuracy reaches 95.9%, and the system has good cross-equipment generalization ability and clinical application value.

The present application relates to a kind of biomarkers, which is selected from one or more of the following: UBR2, NPY4R, KLRF2, HABP2, GLP1R, DR1, RUNX2, ARHGAP5, LINC02135, DDHD1, MIR4523, TC2N.The biomarker is especially used for diagnosing stroke and differential diagnosis of cerebral hemorrhage.The present application screens 12 characteristic biomarkers to construct model, adopts receiver operating characteristic (ROC) analysis to evaluate the performance of model, uses sklearn to calculate area under curve (AUC), the sensitivity and specificity of model reach 0.898, 0.818 respectively, AUC value is 0.913;While differential diagnosis of cerebral hemorrhage, sensitivity and specificity reach 0.864, 0-854 respectively, AUC value is 0.909.The present application combines 5hmC modification spectrum of plasma evDNA and machine learning algorithm for the first time, the characteristic biomarker and / or model screened has the advantages of strong specificity and high sensitivity, overcome the problem of low efficiency and poor accuracy in the process of identifying stroke and cerebral hemorrhage in prior art.

The invention relates to a biomarker. The biomarker is selected from one or more of the following components: UBR2, NPY4R, KLRF2, HABP2, GLP1R, DR1, RUNX2, ARHGAP5, LINC02135, DDHD1, MIR4523 and TC2N. The biomarker is especially used for diagnosing cerebral apoplexy and identifying and diagnosing cerebral hemorrhage. According to the invention, 12 characteristic biomarkers are screened for model construction, the performance of the model is analyzed and evaluated by using the working characteristic (ROC) of a subject, the area under a curve (AUC) is calculated by using sklearn, the sensitivity and specificity of the model respectively reach 0.898 and 0.818, and the AUC value is 0.913; meanwhile, cerebral hemorrhage is identified and diagnosed, the sensitivity and the specificity respectively reach 0.864 and 0-854, and the AUC value is 0.909. According to the invention, the 5hmC modification spectrum of plasma evDNA is combined with a machine learning algorithm for the first time, the screened characteristic biomarker and / or model have the advantages of strong specificity and high sensitivity, and the problems of low efficiency and poor accuracy in the cerebral apoplexy and cerebral hemorrhage identification process in the prior art are solved.

The machine learning-based multi-element data prospecting prediction system comprises a module 1 deposit model construction module, a module 2 spatial database construction module, a module 3 data standardization module, a module 4 prediction model training data distribution module, a module 5 machine learning-based prospecting prediction model construction module, a module 6 algorithm optimization and parameter adjustment module and a module 7 potential area analysis and target area delineation module, and can sequentially perform the following steps: preferential selection of a favorable ore-forming layer, comprehensive processing of multi-element information and establishment of a multi-element spatial database, data standardization and gridding of the database, multi-model multi-parameter ore-forming potential analysis based on artificial neural networks and support vector machines, quantitative evaluation of the prediction model by using an ROC receiver operating characteristic curve and determination of an optimal prospecting prediction model, and determination of a threshold value by using a Youden index and delineation of an ore-forming prospective prediction area or target area distribution map.

This invention proposes a clinical prediction model and design method for the probability of developing anal fistula in Crohn's disease patients, belonging to the field of clinical medical technology. The risk prediction model includes: an input module receiving patient parameters consisting of age at diagnosis, biological agent mode switching, and Montreal manifestation classification at initial diagnosis; a processing module connected to the input module, calculating the assessment and prediction probability of anal fistula in Crohn's disease patients based on the age at diagnosis, biological agent mode switching, and Montreal manifestation classification data at initial diagnosis, according to the following formula; and an output module connected to the processing module, outputting the risk probability of anal fistula in Crohn's disease patients obtained from the risk prediction model. The prediction model is validated on the training and validation sets using the area under the receiver operating characteristic (ROC) curve and a calibration curve to determine its discriminative power and calibration. It can accurately predict the risk of anal fistula in Crohn's disease patients before treatment.

The invention discloses an ovarian cancer diagnosis model based on a machine learning algorithm and a construction method and application thereof, and the construction method comprises the following steps: obtaining external vesicle proteomics data of a clinical ovarian cancer patient and a cell line, and carrying out the normalization preprocessing of the data through Python medium-high throughput analysis; differential protein screening is carried out by utilizing a bioinformatics method, feature selection is carried out in combination with a machine learning algorithm, and nine kinds of external vesicle protein biomarker diagnosis models are constructed; evaluating model performance by using a working characteristic curve of a subject, calculating an area under the curve as a discrimination performance index, optimizing a diagnosis threshold value, and obtaining an optimal model through a ridge regression method; and testing an external test set by using the optimal model, and evaluating the performance of the model by using the working characteristic curve of the subject. According to the diagnostic model disclosed by the invention, the analysis efficiency of exosome data can be remarkably improved, the generalization ability of the model is improved, and the false positive rate is effectively reduced, so that the clinical applicability of the diagnostic model is enhanced.

A system for helping an operator to question a help-seeker is provided. The system includes: taking the positive probability of any second reference group included as a cut-off point, and performing a classification test on any first reference group to obtain a classification sensitivity and a classification specificity corresponding to any second reference group; based on the classification sensitivities and the classification specificities corresponding to a plurality of second reference groups amongst any first reference group, obtaining a receiver operating characteristic curve corresponding to any first reference group; according to the receiver operating characteristic curve, determining a target first reference group from the plurality of first reference groups; and according to the distance between a coordinate point corresponding to any second reference group amongst the plurality of second reference groups and a perfect classification coordinate point, determining a target question semantic meaning.

The present application relates to the technical field of medical information processing and machine learning, and particularly relates to a cardiovascular disease prediction method based on cross combination strategy and dynamic weighted stacking. The method comprises the following steps: S1, collecting data and constructing a data set; S2, constructing a heterogeneous base model set according to the data set; S3, taking the area under the receiver operating characteristic curve as an evaluation index, and selecting 10 heterogeneous base models with the optimal performance; S4, assigning adaptive weights to the 10 heterogeneous base models selected in S3; S5, using the adaptive weights to weight and fuse the prediction probabilities of the 10 heterogeneous base models, and generating a meta-feature data set for training a meta-model; S6, selecting a meta-model, and training the meta-model using the meta-feature data set; S7, inputting data into the trained meta-model, and obtaining a prediction result according to the meta-model; and S8, analyzing the influence of each feature on the risk of disease. The present application solves the problem of insufficient feature correlation mining through full cross combination of multi-feature selection methods and multi-classifiers.

The invention relates to the technical field of medical intelligence, in particular to a chronic obstructive pulmonary disease early screening model construction method based on machine learning and multi-dimensional indexes. Collecting multi-dimensional data; eliminating dimensional influence on the multi-dimensional data based on standardization processing; screening indexes of the standardized multi-dimensional data on the basis of a Logi st ic regression algorithm; inputting the screened indexes into a multi-statistic channel attention one-dimensional convolutional neural network algorithm to construct a chronic obstructive pulmonary disease early screening model; training the chronic obstructive pulmonary premature screening model based on the training set, evaluating the chronic obstructive pulmonary premature screening model by using the test set, and adjusting model parameters according to an evaluation result until the performance of the chronic obstructive pulmonary premature screening model is optimal; and evaluating the chronic obstructive pulmonary premature screening model based on the confusion matrix, the subject working characteristic curve (ROC) and the area under the curve (AUC). The problems of low accuracy and low efficiency of early screening of the chronic obstructive pulmonary disease in the prior art are solved.

The application discloses a liver cancer risk data processing method and system based on machine learning, comprising the following steps: acquiring a sample data set containing multi-dimensional clinical test indexes; screening a cross-algorithm consistency feature set through multi-algorithm cross-validation; constructing multiple machine learning models in parallel; evaluating and verifying the models by using a receiver operating characteristic curve, a calibration curve and a decision curve, and screening an optimal model; and verifying the prediction performance of the model on different data subsets. The data processing method adopted by the application has good classification performance and generalization ability by deeply mining data features through a machine learning algorithm based on conventional clinical test indexes.

The present invention discloses a method for assessing traffic accident risk in a continuous flow road scenario, comprising: constructing a case-control dataset, dividing it into historical sample data and test data, performing self-organizing map cluster analysis using the historical sample data, and building a risk scenario identification model; inputting a test dataset, determining the risk scenario to which the test data belongs, training a base learner model using the historical sample data of the risk scenario as a training set, and outputting prediction results for the test data; plotting a receiver operating characteristic (ROC) curve for the prediction results of the test data, evaluating goodness of fit using the area under the curve (AUC) metric, and selecting an optimal risk threshold based on the Youden index; obtaining traffic flow parameters on the road to be assessed in real time, determining the risk scenario, and calculating the risk level on the road to be assessed in real time. The present invention provides accurate and real-time early warning of road traffic accident risks at a relatively low cost, thereby improving the safety and reliability of traffic system operation.

The present application belongs to the technical field of molecular medicine, and relates to a probe, a marker combination and a prediction method for non-invasive risk grading of neuroblastoma. The probe UUU-DZ-tFNA can detect UDG in vitro, cells, living bodies and plasma exosomes, and it is proved that UDG in plasma exosomes is closely related to the risk grading of NB. The plasma exosomes of NB patients are collected MYCN Whether to amplify, whether the tumor is metastatic, neuron-specific enolase content, lactate dehydrogenase content and plasma exosome UDG content are integrated, and a NB non-invasive risk grading prediction model is established through a machine learning algorithm. The optimal risk grading model is a neural network model. The area under the receiver operating characteristic curve of the combined model is improved by 5.2% compared with the optimal prediction result of the combined prediction model of the single clinical index, and the sensitivity and specificity are significantly improved. The present application has key clinical values for accurate grading of NB children, individualized treatment plan formulation and efficacy monitoring.

The invention discloses a kidney transplantation rejection reaction subtype classification system based on artificial intelligence, which relates to the field of kidney transplantation and comprises a data collection module, a feature aggregation and classification module, a feature extraction module and a model evaluation module. The system is used for extracting pathological image feature vectors and clinical feature vectors according to the collected pathological image data and the clinical feature number, and then splicing and classifying the pathological image feature vectors and the clinical feature vectors to obtain a subtype classification result of the kidney transplantation rejection reaction; and a classification result is evaluated through a confusion matrix, a receiver operation characteristic curve ROC and an area under the curve AUC so as to ensure the accuracy and reliability of the result. According to the invention, the accuracy and repeatability of diagnosis can be improved, the workload of pathologists can be reduced, multi-modal information can be more fully integrated, and the comprehensiveness and precision of diagnosis can be improved.

The application discloses a synthetic aperture radar image adversarial sample detection system and method based on multi-target optimal feature attribution selection. Historical SAR images are collected from a SAR system monitoring database, and after data standardization and normalization, the historical SAR images are used as input data sets. The number of sub-samples generated by a feature analysis process based on sliding scanning and the area under the receiver operating characteristic curve of a logistic regression model are used as optimization targets. A feature scanning block parameter optimization platform based on a multi-target optimization method is designed to obtain optimal feature attribution scanning blocks and an optimal regression model. The optimal regression model is used for online detection of real-time SAR image data in a real-time database of a SAR system. The application can automatically obtain the best feature analysis granularity according to different scenes, efficiently realize various adversarial sample detections in the SAR image recognition field, and improve the calculation efficiency and AUC performance index of the SAR image adversarial detection.

The data compression analyzer can quickly make a binary decision to compress or not compress an input data block, or it can use a relatively slow neural network to predict the compression ratio of the data block through a regression model. The concentration (CV) and the number of zero-value (NZ) symbols are calculated based on the sum of squared frequencies from an unsorted symbol frequency table. A compression decision is made quickly when their product, CV*NZ, exceeds a level threshold THH. During training, CV*NZ is plotted as a function of the compression ratio C% for multiple training data blocks. Different THH test values ​​are applied to this plot to determine the true positive rate and false positive rate, and are plotted as a receiver operating characteristic (ROC) curve. The point on the ROC curve with the highest Yoden index is selected as the optimal THH for future binary decisions.

The invention belongs to the field of electroencephalogram signal processing, nerve decoding, neuroinformatics and sensory stimulation recognition, and provides a dynamic time sequence decoding method for electroencephalogram signals under sensory stimulation, which comprises the following steps of: preprocessing the electroencephalogram signals acquired in different sensory stimulation processes; then slicing the preprocessed electroencephalogram data by time points, and constructing an instantaneous feature matrix; training a classifier on each time slice by adopting a support vector machine and calculating the area under the working characteristic curve of the subject; and recognizing a time window of which the classification performance is remarkably higher than a random level by combining replacement test based on clusters, so as to obtain characteristic nerve response modes induced by different sensory stimuli. According to the method, an electroencephalogram time sequence decoding method is introduced into sensory stimulation recognition for the first time, the nerve response process can be analyzed on the millisecond scale, a stimulation specific significant time window can be extracted, a universal stimulation recognition template is constructed, and automatic, objective and quantifiable sensory stimulation classification is achieved.

Embodiments of the present application provide a system for guiding an operator to question a rescue sender. The system uses the positive example probability of any second reference group included in any first reference group among a plurality of first reference groups as a cut-off point to perform a classification test on any first reference group to obtain the classification sensitivity and classification specificity corresponding to any second reference group, and based on the classification sensitivity and classification specificity corresponding to a plurality of second reference groups within any first reference group, obtain a receiver operating characteristic curve corresponding to any first reference group. Based on the receiver operating characteristic curves corresponding to each first reference group within a plurality of first reference groups, determine a target first reference group from the plurality of first reference groups, and determine target question semantics from the distance between the coordinate point corresponding to any second reference group among the plurality of second reference groups and the perfect classification coordinate point.

The present invention discloses a method based on trans m 6 A regulatory factor expression profile and cis-m 6 A sequence combination feature predicts m in single cells 6 A methylation profiling method, including the establishment of single-cell m 6 A methylation profile prediction model and single-cell m 6 A methylation profile prediction. The batch prediction model constructed by this invention has good fitting effect. The difference between the true value and the predicted value is defined as true when it is within the range of -0.5 to 0.5, and false when it is outside the range. When the receiver operating characteristic (ROC) curve is plotted, the average area under the curve (AUC) of 4162 models reaches 0.91. The prediction speed is also fast, and the prediction time for samples containing tens of thousands of single cells only takes a few minutes.