16 results about "Multiple classifier" patented technology

The invention discloses a multi-endpoint reproductive toxicity prediction method and system based on a large language model, and belongs to the technical field of toxicity prediction. The method comprises the following steps: extracting phenotype information of text data in a candidate data set by using the large language model; performing multi-dimensional annotation processing on the phenotype level annotation data set; inputting the reproductive toxicity training set into a pre-constructed molecular representation learning model for training, and constructing a reproductive toxicity prediction model set with phenotype specificity; performing performance evaluation on the plurality of classifiers in the reproductive toxicity prediction model set by using the test data set, and screening according to preset performance indexes to obtain an optimal prediction model combination; performing toxicity prediction on the molecular structure of the chemical substance to be detected by using the optimal prediction model combination to obtain a multi-dimensional reproductive toxicity prediction result including cross-gender, cross-generation and cross-research types; according to the method, high-resolution and multi-dimensional reproductive toxicity phenotype prediction is carried out on the environmental chemical substances.

The invention discloses a surge absorber early failure diagnosis method based on acoustic characteristics. The surge absorber early failure diagnosis method comprises the steps of collecting acoustic signals generated by a surge absorber in an operation state; performing mixed feature extraction by fusing improved variational mode decomposition and frequency cepstrum coefficients, and constructing a high-dimensional initial feature set; pre-screening the high-dimensional feature set by using a minimum redundancy and maximum correlation algorithm, and then performing feature fine optimization by using a support vector machine package method optimized by a quantum behavior particle swarm algorithm to obtain an optimal feature subset; and finally, constructing a plurality of classifier groups based on self-service sampling, and performing decision-making layer fusion on the output of each classifier by using a D-S evidence theory to realize accurate diagnosis of the early failure state of the surge absorber. Through deep fusion of multiple levels and multiple algorithms, the sensitivity, robustness and accuracy of diagnosis are remarkably improved, and non-intrusive online monitoring and early warning of the surge absorber can be realized.

This invention discloses a small-sample fault diagnosis method based on multi-scale ensemble LightGBM, comprising the following steps: acquiring vibration acceleration signals of rotating machinery and extracting samples to obtain training and test sets; constructing a multi-scale ensemble model consisting of a data augmentation module, a label classification module, and an output weighting module; in the data augmentation module, resampling the original samples using a multi-scale sliding window to obtain multiple sub-sample training and sub-sample test sets; in the training phase, training multiple classifiers in the label classification module using the sub-sample training sets; in the testing phase, inputting the sub-sample test sets into the corresponding classifiers, and obtaining the test results of the label classification module on the original samples through ensemble; calculating the frequency domain similarity between the test set and the training set samples using a metric function, and using it as a weight; in the output weighting module, weighting the ensembled test results to obtain the final classification diagnosis result.

This invention discloses a method and system for identifying the spatial distribution of accumulated water based on multi-classifier collaboration. The method includes: constructing a hierarchical classifier library comprising a preliminary screening layer, an arbitration layer, a re-identification layer, and a verification layer, with each layer containing multiple classifiers of different types. First, the preliminary screening layer performs coarse classification on UAV images, and the results of each classifier are fused using a logical OR operation to determine suspicious areas. Then, the arbitration layer and the re-identification layer sequentially perform refined identification of ambiguous areas, using a logical AND operation to obtain high-confidence accumulated water areas. For the remaining ambiguous areas, the verification layer is activated for final discrimination. The high-confidence accumulated water areas identified by each layer are merged to form a preliminary accumulated water distribution map. Pixel confidence is optimized through weighted fusion, and low-confidence pixels are removed. Finally, post-processing is performed to output the final spatial distribution result of the accumulated water. This application can effectively improve the identification accuracy and robustness of small and irregular accumulated water areas in complex urban scenes.

This invention discloses a classification method for expanding training samples based on multi-classifier recognition results, comprising: selecting multiple classification methods; selecting initial training samples including each category from the dataset to be classified to form an initial training sample set; classifying and recognizing the dataset data using each classification method and the training sample set to obtain the classification results of the dataset to be classified by each classification method; calculating the classification result acceptance rate for each data based on the classification results of the dataset to be classified; comparing the classification result acceptance rate with its preset threshold to obtain new training samples and expand the training sample set; determining and executing the next iteration classification based on the expanded training sample set; and taking the maximum acceptance classification result of the last iteration as the final data classification result of the dataset to be classified. This invention improves classification accuracy by iteratively verifying the results of multiple classifiers and gradually expanding the training samples.

The invention discloses an underwater sound source target identification method based on a dynamic selection integration technology, and the method comprises the steps: obtaining a sound source signal in real time through an underwater sensor, including acoustic features of a submarine target and a sound bait, extracting a time-frequency feature, a modulation feature and a statistical feature through short-time Fourier transform, and fusing the features through PCA (principal component analysis). The method comprises the following steps: generating a multi-dimensional feature vector, establishing a plurality of classifier models, evaluating the performance through cross validation, intelligently selecting an optimal classifier combination, inputting the fused feature vector into the optimal combination for target classification, outputting an identification result, and finally, continuously updating the classifier models by adopting an incremental learning method according to the identification result and an actual situation. And continuous optimization management of underwater sound source target identification is realized. Through a multi-dimensional feature fusion mechanism, time-frequency features, modulation features, statistical features and other information are comprehensively considered, the distinguishing capacity of the model for a target and an interferent is improved, in addition, through intelligent selection of an optimal combination in multiple classifiers, the distinguishing capacity of a submarine target and a sound bait is improved, and the recognition precision is effectively enhanced.

The invention discloses a cross-user wearable activity identification method based on group specific concept perception representation learning. The method comprises the steps of collecting and preprocessing multi-user sensor data; the concept offset degree is measured from a time sequence view angle and a semantic view angle, user clustering is carried out by fusing multi-view-angle measurement results, and group specific concept tags are generated; constructing a perception model comprising an activity encoder, a user encoder and a plurality of classifiers, and performing supervised learning by minimizing the joint loss of activity, user and group specific concept classification; introducing a condition discriminator to construct a representation pair of joint distribution and edge distribution, and minimizing condition mutual information of activity representation and user representation under a group specific concept through adversarial training to obtain a trained model; during application, test data are input into the trained model for activity identification. According to the method, through explicit modeling of specific concepts and decoupling of activity and user features, cross-user concept offset is eliminated, and the activity identification generalization ability of the model on new users is significantly improved.

The application discloses a strip plate shape intelligent regulation and control method based on an adaptive fusion model, which comprises the following steps: step 1, collecting historical rolling production process data and performing pretreatment; step 2, constructing a rolling production process data set based on the pretreated data; step 3, classifying plate shape quality according to the ratio of strip crown and target thickness and setting a category label; step 4, establishing an adaptive strip outlet plate shape diagnosis model containing multiple classifiers based on a DS theory, and training the diagnosis model through the rolling production process data set; step 5, inputting rolling production process data under a new rolling schedule into the classifier of the trained adaptive strip outlet plate shape diagnosis model to obtain a real-time classification prediction result of the strip, and if the prediction classification is under-crown or over-crown, step 6 is executed; otherwise, the process parameters are not adjusted; and step 6, dynamically adjusting and optimizing the process parameters based on an inverse linear quadratic control (ILQ) according to the prediction result of step 5.

The application discloses a cross-user wearable activity recognition method based on group-specific concept-aware representation learning, which comprises the following steps: collecting multi-user sensor data and preprocessing; measuring the concept drift degree from the time sequence perspective and the semantic perspective respectively, fusing the multi-perspective measurement results to perform user clustering, and generating group-specific concept labels; constructing a perception model comprising an activity encoder, a user encoder and multiple classifiers, and performing supervised learning by minimizing the joint loss of activity, user and group-specific concept classification; introducing a conditional discriminator to construct a representation pair of joint distribution and marginal distribution, minimizing the conditional mutual information of activity representation and user representation under the group-specific concept through adversarial training, and obtaining a trained model; and inputting test data into the trained model for activity recognition. The application explicitly models the group-specific concept and decouples the activity and user features, eliminates the cross-user concept drift, and significantly improves the activity recognition generalization ability of the model on new users.

The invention discloses a cross-domain target detection method based on characteristic decomposition, a computer program product, a storage medium and a terminal, and belongs to the technical field of target detection. The method comprises the following steps: generating an intermediate domain image which retains high-level semantic features of a source domain and a target domain through an intermediate domain generator; fusing the last three layers of features of the backbone network through a multi-scale feature fusion module; a feature decomposition module is introduced, the module comprises a global decoupling module and an instance-level decoupling module, and consistent information representation between a source domain and a target domain is achieved through feature decoupling and utilization; and setting a subtask alignment module which comprises a plurality of classifiers and locators as auxiliary predictors and is used for adjusting the feature representation of the intermediate domain and the target domain and realizing the decoupling optimization of classification and positioning tasks. According to the method, the problems of insufficient context information utilization, feature representation coupling and the like in the prior art are effectively solved, the performance in multiple cross-domain scenes is improved, and the average precision is improved by about 1.5%.

The application provides a citation network node classification method and system fusing GCN and multiple classifiers, wherein the classification method comprises the following steps: regarding the whole citation network as a non-directional graph G comprising node attributes and topological structures; calculating an adjacency matrix and a feature matrix of the non-directional graph G; initializing an edge weight matrix of the non-directional graph G, and calculating the edge weight of the non-directional graph G according to the feature matrix; converging the calculated edge weight to the nodes at both ends of the edge, updating the feature matrix; filtering and amplifying the features in the updated feature matrix; inputting the adjacency matrix and the updated feature matrix into n classifiers respectively for prediction and fusing the prediction results to obtain the final classification result. The classification method comprehensively considers the effective semantic information in the network, can obtain more accurate and stable classification effect through the prediction of multiple classifiers and the fusion by using fuzzy integration.

The invention discloses an unbalanced sample-oriented alarm flooding data enhancement and root diagnosis method, which comprises the following steps of: acquiring alarm event data generated in an industrial system; the improved SMOTENC algorithm performs data enhancement on minority class samples, and a synthetic alarm sample having highly similar sequential logic with an original fault sample is generated in a feature space to balance a data set; constructing a group of hidden Markov model variant classifiers with different structures; evaluating the performance of each variant classifier by adopting different indexes; and taking different performance indexes as references, fusing diagnosis results of a plurality of classifiers by adopting an integrated learning framework of a weighted voting method, and finally outputting an alarm source diagnosis result with high confidence in real time. By improving the data enhancement method, the problem of sample imbalance in the industrial alarm data can be effectively solved; the advantages of different HMM variants are integrated through an innovative model integration framework, and the accuracy of alarm flooding root diagnosis can be remarkably improved.

The invention provides a remote sensing image scene classification method and device based on a multi-stream self-distillation architecture. The multi-stream self-distillation network architecture composed of a main stream and two branch streams is constructed. The main stream is constructed based on a CNN model, and the two branch streams are constructed based on a MobileNetV2 model. Attention modules are introduced into the two branch streams to enhance feature interaction with the main stream. In addition, a feature fusion strategy and a classification result integration strategy are introduced into the architecture to construct a feature fusion classifier and an integrated classifier. During architecture training, a self-distillation strategy is adopted, and an integrated classifier is used as a teacher model to guide training of each branch model, a trunk model and a feature fusion classifier. Experiments on four remote sensing image data sets show that the multi-stream self-distillation architecture can improve the classification precision on the basis of not changing the CNN network structure, and output multiple classifiers with different precision and resource consumption levels to adapt to multiple practical application scenes such as edge calculation and high-precision recognition.

The control chart pattern recognition method based on weighted ordered patterns and ensemble classifiers includes: Step 1: Collecting partial data of the controlled process in actual manufacturing, and generating eight abnormal control chart patterns (CCPs) using the Monte Carlo method; Step 2: For CCP observation data of a given length, reconstructing it into a symbolic phase space according to a specific parameter combination; Step 3: Obtaining all ordered pattern types based on the embedding dimension parameter, and calculating the ordered pattern features of the identified CCPs; Step 4: Generating weighted ordered pattern (WOP) features, calculating the amplitude features of the identified CCPs, and weighting the ordered pattern features from Step 3 to obtain WOP features; Step 5: Changing the time delay parameter to obtain WOPs corresponding to different time delay parameters, constructing a homogeneous classifier with the same number of time delay parameters, and performing individual classifier classification; Step 6: Based on a democratic voting method, voting on the recognition results of multiple classifiers in Step 5 to obtain the comprehensive diagnostic result of the identified CCPs.

An entity relation extraction method based on pre-judgment of token sequence and multi-round classification, a model for predicting text sentences at a token sequence, i.e., span level, is used to find all entities and the relations between entities in the text sentences. The model uses a BERT pre-training model and simultaneously includes three modules, i.e., pre-judgment (PEJ), entity multi-round classification (EMR) and relation multi-round classification (RMR). Through the preliminary judgment of the PEJ module and the multi-round entity classification of the EMR module, entity recognition is performed, and then the RMR module is used for multi-round relation classification to determine the relations between entity pairs, thereby realizing relation extraction. The multi-round classification in the Smrc model enables the data set to be used multiple times and sufficiently, the model is better fitted, multiple classifiers directly determine the relations between entities and entity pairs in multiple categories, the problem of imbalance and large difference in discrimination ability of a single multi-output classifier in different categories is avoided, the model structure is more refined, and the entity pre-judgment module (PEJ) is added, so that the Smrc model has more accurate recognition effect.