37 results about "Bootstrap aggregating" patented technology

The present invention discloses a selective integrated learning-based rotating machinery fault diagnosis method. The method comprises the steps of pre-processing a known input signal to obtain a training sequence; based on the Bagging algorithm, processing the training sequence to generate a series of differentiation-based learning machines; through the selective integrated learning process, giving a preference on the series of differentiation-based learning machines to obtain a well trained PSOSEN model; by utilizing the trained PSOSEN model, diagnosing a rotating machinery fault. The method solves the problem in the prior art that a classifier with better performances is difficult to be selected out of a series of classifiers can be solved.

The invention discloses a lithium ion battery service life forecasting method based on an integrated model and relates to a lithium ion battery cycle life forecasting method. The lithium ion battery service life forecasting method is used for solving the problem that the existing lithium ion battery is low in service life forecasting adaptability and poor in stability. The lithium ion battery service life forecasting method includes: performing preprocessing on battery cycle charging and discharging test testing data; adopting a Bagging algorithm to perform secondary resampling on a Train database; building a monotonous echo state network model; initializing inner connection weights of a monotonous echo state network, and repeating for T times to obtain T untrained monotonous echo state network sub-models; setting a first free parameter set and a second free parameter set of the monotonous echo state network model; integrating output RULi of the monotonous echo state network model, adopting the Test database to drive the integrated monotonous echo state network model, and obtaining remaining service life of a lithium ion battery. The lithium ion battery service life forecasting method based on the integrated model is suitable for lithium ion battery service life forecasting.

The invention discloses a fatigue detection method based on multi-source information fusion. Electroencephalogram signals, twinkling information and electrocardiosignals of a testee are synchronously collected by means of an electroencephalogram collecting device and an electrocardiogram collecting device respectively; electroencephalogram signal features including the relative energy of electroencephalogram rhythm waves alpha, beta, theta and delta, electro-oculogram information including twinkling frequency E and twinkling intensity F, and electroencephalogram features including heart rate values HR, LF and HF are extracted; by means of the logistic regression algorithm, the fatigue degrees are primarily divided into three classes, namely, the non-fatigue degree, the mild fatigue degree and the deep fatigue degree, and meanwhile features with large weights are screened according to logistic regression weights for feature fusion; fused feature vectors are classified again by means of the bagging algorithm based on a support vector machine, the processed feature vectors serve as input of the bagging algorithm, and the current fatigue degree of the testee is determined; different fatigue relieving methods are used according to classification results of the fatigue degree of the testee. The method has the advantages of being high in applicability, high in fatigue detection precision, good in improvement effect and the like.

The present invention discloses a lake-reservoir cyanobacteria water bloom recognition method based on remote sensing monitoring and evidence fusion technology improvement, and belongs to the technical field of environmental projects. The method comprises: establishing a remote sensing inversion model, determining cyanobacteria water bloom breaking-out degree recognition indexes, pre-treating monitoring data, assigning belief function values, and recognizing monitoring area cyanobacteria water bloom. According to the present invention, an ensemble learning Bagging algorithm is fused into regression analysis of the remote sensing inversion model (ie., the normalization vegetation index model) to correct parameters of the model to obtain the corrected remote sensing inversion model so as to improve fitting accuracy; the problem of selection of indexes for lake-reservoir cyanobacteria water bloom recognition and value ranges thereof is solved; the problems of selection of the monitoring area and pretreatment of the monitoring data are solved; the problem of assignment of the belief function values is solved; and effective cyanobacterial water bloom recognition is achieved.

The invention belongs to the field of underwater robot control, and discloses a method for planning paths on the basis of ocean current prediction models. The method includes steps of carrying out rasterization processing on sailing regions according to path critical points; predicting ocean current for the sailing regions by the aid of regional ocean modes and acquiring real-time ocean current information by means of fitting computation; marking prohibited areas by the aid of electronic ocean map information; storing prohibition information of different depths and starting point and end pointlocation information according to plain grids at different depths, and storing longitudes and latitudes of various points of the grids, whether the points are the prohibited areas or not and whetherend points are reached or not; computing the directions from current locations to the end points and determining optional actions in travel directions at all next steps; seeking the optimal strategiesby the aid of Q-learning and outputting the paths. The optimal strategies are planned in Markov decision-making processes. The method has the advantages that influence of the real-time ocean currenton path planning is sufficiently considered, fitting is carried out by the aid of BP (back propagation) neural networks and bagging algorithms, the optimal solution can be sought by means of reinforcement learning, accordingly, the convergence speeds can be increased, and the computational complexity can be lowered.

The invention discloses a method of using deep learning for movie recommendation based on emotion analysis. Movie online-comments of a user and movie status information are acquired, and cleaning operations and pre-processing are carried out on movies and comment information thereof; structure and feature information extraction of data is carried out, and emotion analysis is carried out on featureinformation through deep learning; and then a user interest model is established on the basis of emotion analysis data, then interest degrees of the user on movies are calculated, and movie recommendation is carried out. According to the method, a Bagging algorithm and a bi-directional long short-term memory (Bi-directional LSTM) network are used as a method of emotion inclination analysis, subjective emotion expressed by the user in the text information is mined, the effective user interest model is constructed, accuracy of a recommendation system is improved, and quality of movie recommendation is optimized.

The invention discloses a river water quality prediction and an evaluation method of water quality influencing factors. The method comprises the following steps: firstly, extracting historical data of river water quality and watershed characteristics of corresponding sampling points to form an original training set; Secondly, through the bagging algorithm, randomly selecting samples from the original training set, and constructing several sub-training sets . third, generating a decision tree by selecting split attribute according to that characteristics of different watersheds, and constructa random forest model according to the decision tree; Fourth, evaluating the simulation effect of the model; 5, acquiring watershed characteristic data of a point to be predicted, putting that data into a stochastic forest model, and obtaining corresponding water quality prediction data; Sixth, assessing the impact of different basin characteristics on river water quality. The relationship modelbetween watershed characteristics including watershed hydrology, climate, geographical characteristics, seasonal factors, human influence and the like and river water quality is built, river water quality index data is accurately predicted according to watershed characteristics of a target point and importance of the influence of a watershed characteristic on water quality is evaluated.

The invention discloses an electricity fee sensitive user analysis method based on stacking and bagging algorithms. The electricity fee sensitive user analysis method comprises steps of using an optimal zone algorithm to solve problems of nonequilibrium and intolerance and construct a training set and a testing set according to electricity fee sensitive classification targets and on the basis of multiple core business indexes of an electricity fee sensitive original data table, further subdividing related core business structuralized characteristics based on electricity fee sensitivity, using a stacking method to construct an electricity sensitive primary model on the basis of nonstructural text characteristics of particles and word frequency statistics, generating extended stacking characteristics for each sample, combining characteristics of the two as an integral input, using a bagging algorithm and a vote algorithm to construct an electricity fee sensitivity secondary model on the training set and the testing set, using a trained model to perform prediction on a verification set and verifying during practical business. The Electricity fee sensitive user analysis method based on stacking and bagging algorithms can improve mastery degree of an electric power company about electricity fee sensitivity of the user, and is beneficial for providing differential and specific quality power supply service.

The invention relates to an unbalanced data classification method based on boundary upsampling. The unbalanced data classification method includes that Euclidean distance is used as the distance metric for each positive sample in the unbalanced dataset, K sample points nearest to the positive samples in the unbalanced dataset are taken, the class labels of the samples are compared, the number ki of negative samples in the K sample points is given, if ki >=K/2, the positive sample point is near to the real decision boundary of positive and negative samples, conversely, the positive sample point is far from the boundary, one positive sample is randomly selected from the K nearest positive samples, and a new positive sample is generated between the selected positive sample and the positive sample selected from the K nearest positive samples, this process is subjected to iterative execution for many times until the positive and negative samples are same in number, and thus a balanced dataset is constructed, and the generated balanced dataset is trained by a Bagging algorithm to obtain a final classification model. According to the invention, a better classification effect can be achieved in the unbalanced dataset.

The invention discloses a soft measurement modeling method based on semi-supervised ensemble learning, belongs to the field of modeling and soft measurement in a complex industrial process, and is used for a chemical process with a small quantity of label samples. The method is an online prediction strategy based on semi-supervision. A Bagging algorithm is adopted to divide a no-label sample set into three sub sample sets, and the label sample is adopted to train three regression models; then, on the basis of a confidence coefficient index, the corresponding index value of the no-label sampleis calculated, the no-label sample which meets a confidence coefficient requirement is selected for carrying out labeling, and the labeled sample is added into the corresponding label sub sample set;and finally, a Gaussian process regression model is independently established for three enlarged label datasets, and a weighting method is adopted for carrying out result fusion. By use of the method,the no-label sample information in the chemical process can be effectively utilized to accurately predict a key variable so as to improve product quality and lower production cost.

The invention discloses an arrhythmia identification and classification method based on sparse representation and neural network. The method comprises the following steps: preprocessing electrocardiosignals by utilizing a sparse representation frame to acquire a low frequency part of QRS (Quality Rating System) wave; subsequently, realizing feature extraction by utilizing discrete cosine transform; carrying out analysis of main components to acquire transform coefficients as feature attributes after dimension reduction; and finally, carrying out automatic classification on heat beats of the six types including normal cardiac rate (N), left bundle branch block (LBBBB), right bundle branch block (RBBBB), auricular premature beat (APB), premature ventricular contraction (PVC) and pacemaker hear beat (PB) in arrhythmia by utilizing a Bagging algorithm using BP (Back Propagation) neural network as a base learner. According to the arrhythmia identification and classification method disclosed by the invention, feature extraction is carried out from a low frequency range of the QRS wave, so that the dimensionality of the feature attributes is reduced; and the problem of unbalance classification is solved by utilizing the Bagging algorithm in ensemble learning, so that the classification accuracy is improved.

The invention provides a lithium ion battery remaining service life prediction method based on an AR ensemble learning model, relates to a lithium ion battery remaining service life prediction method and aims at solving the problem that an existing single AR model is limited in accuracy and low in stability during nonlinear time series prediction. The method performs prediction on the lithium ion battery remaining service life based on the AR ensemble learning model, vectors composed by input data are randomly selected to form a group of sub-vector sets by adopting a Bagging (Bootstrap Aggregating) integration method, one AR model is input into each vector set to perform parameter computation and capacity prediction, finally prediction results are integrated and output, a capacity degeneration curve and a probability density curve are drawn, and finally a final prediction output is obtained. The lithium ion battery remaining service life prediction method can improve the stability and accuracy of the lithium ion battery remaining service life prediction and is suitable for lithium ion battery remaining service life prediction.

The detection method includes generating a plurality of neural network models. Each model has as a training set a data set from a plurality of samples of a commodity of known origins. Each sample has been analyzed for a plurality of elemental concentrations. Each neural network model is presented for classification a test data set from a plurality of samples of a commodity of unknown origins. As with the training set, the samples have been analyzed for the same plurality of elemental concentrations. Next a bootstrap aggregating strategy is employed to combine the results of the classifications for each sample in the test data set made by each neural network model. Finally, a determination is made from the bootstrap aggregating strategy as to a final classification of each sample in the test data set. This final classification is indicative of the geographical origin of the commodity. The system includes software for generating the neural network models and a software routine for performing the bootstrap aggregating strategy.

The invention belongs to the technical field of electroencephalogram signal processing. The invention discloses a hierarchical Bagging method for sentiment analysis based on electroencephalogram signals. The hierarchical Bagging method comprises the steps of electroencephalogram sample data preprocessing, feature extraction and feature selection, replacement sampling of a training set, training ofa plurality of data subsets through different base classification algorithms and voting of a plurality of classifiers to obtain a classification result. Different from a traditional Bagging algorithmin which a single training subset corresponds to a single classification algorithm, hierarchical Bagging enables a plurality of training subsets to correspond to a single classification algorithm, and the risk that a classification algorithm with good single performance is deleted due to the fact that the classification algorithm does not adapt to individual data is reduced. According to the method, the accuracy of electroencephalogram signal classification can be effectively improved, the problem of poor stability of a single classification algorithm is solved, and the method can also be popularized to other similar types of data processing. The method is of great significance to emotion monitoring, risk prediction and supervised learning classification.

The invention discloses a data-driven SPI defect type intelligent identification method on an SMT production line. The method comprises the following steps of in a first stage, carrying out clusteringprocessing on SPI historical quality detection data sets, independently sampling the obtained K types of training data sets for 20 times by adopting a Bagging algorithm, and respectively training independent defect classifiers for K types of K * 20 groups of training sets by utilizing a BP neural network model to obtain K * 20 independent defect classifiers to form a classifier set; in A second stage, detecting 6 solder paste printing quality parameters online; comparing with a historical training data set to classify the detection records T; determining which category of the real-time detection point belongs to the K categories of training data sets, and when T is just located on the boundary of two or more categories of training data sets, simultaneously selecting twenty independent defect classifiers from the multiple categories of K categories of training data sets according to approximately equal quantity to carry out category judgment of detection records T; And inputting T intoeach independent defect classifier, carrying out integrated prediction on an output result according to an integration rule, and judging a defect type. According to the invention, the effect of people in automatic detection is reduced, and the online real-time detection efficiency and accuracy are improved.

The invention discloses an improved algorithm for missing value interpolation. The algorithm comprises the steps that hierarchical clustering is performed on all data; for a class containing missing values, according to the judgment of whether a record of the missing values is available, the record is divided into a complete data set m1 and a missing data set m2; the data in m1 is randomly dividedinto a training set and a test set, k types of interpolation methods are used to predict the test set, and the method with the highest accuracy is obtained; whether the obtained method is a weak method is judged, if yes, a function for missing value interpolation of the class is obtained in combination with a bagging algorithm, and if not, the algorithm is a final algorithm; the final interpolation function is adopted to perform interpolation on the missing values of the class; and whether another class containing missing values exists is judged. According to the improved algorithm for missing value interpolation, the method suitable for the data set can be selected from numerous missing value interpolation schemes according to specific properties of the data, interpolation effects of themissing values through various methods are compared according to the principle of the bagging algorithm, and therefore the practical method for missing value interpolation of the data is obtained.

The invention relates to an electroencephalogram (EEG) recognition method. The method comprises the following steps: carrying out a character experiment to extract an EEG including 300 signals so as to serve as a training set; integrating a plurality of support vector machines with mixed kernels to serve as a learner by utilizing a bagging algorithm, and adaptively adjusting parameters of the learner based on an immune algorithm by adopting the training set so as to obtain optimum parameters; and finally, recognizing a P300 signal in the EEG by utilizing the learner with the optimum parameters, wherein the optimum parameters refer to parameters capable of enabling the learner to accurately recognize the P300 signal, and accurate recognition is that the accuracy of repeated experiments of more than 12 times is 96-98%. According to the EEG recognition method disclosed by the invention, the parameters can be intelligently selected according to optimized contents, the defects that the traditional learner needs to be continuously adjusted and optimized and needs a cross validation process are overcome, the intelligence of the integrated learner is improved, and the EEG recognition method disclosed by the invention is excellent in recognition performance, high in accuracy rate and high in overall generalization ability and has excellent popularization and application values.

The invention discloses an improved LR-Bagging algorithm based on feature selection, which comprises the following steps: first, determine the initial data set from the original data, and require that the degree of correlation between the independent variable and the dependent variable cannot be too low; secondly, the initial data set Discrete independent variables are encoded by WEO; then random sampling is used to obtain a certain number of records and feature fields to form training examples, and the training examples are trained with LR ((LogisticRegression) model and the normal significance test of the coefficient is performed. If not significant, then Eliminate, on the contrary, add the combination model. Carry out cyclic iterations until the combination model is better. Finally, you can use the better combination model for prediction and grouping. This algorithm can improve the diversity of classification results, the degree of extraction of variable information and the prediction results It can also effectively reduce the possibility of multicollinearity and "overfitting" caused by too many variables in the base LR model.

The invention provides an electric vehicle charging facility fault prediction method and system, and the method comprises the steps of reading a charging data set, and carrying out the type division of the data set; setting the data range of each hyper-parameter; selecting a group of hyper-parameters with the highest accuracy as hyper-parameters of the fault prediction model to reinitialize the model; generating different sub-data sets by using a Bagging algorithm; respectively submitting different sub-data sets to the corresponding decision tree models for regression analysis; and according to different output weights of each decision tree during training, outputting a result obtained after all outputs are unified as distance fault prediction time of the model. According to the method, the fault prediction of the charging facility is realized, the prevention measures can be taken before faults really occur, the faults are avoided, the part damage, service interruption and the like caused by the faults are reduced, and the facilities are prevented from entering an unsafe or uncertain state caused by the faults, so that the operation and maintenance cost is reduced, the equipment operation efficiency is improved, and the safety is ensured.

The invention discloses a fatigue driving detection system and a use method thereof. The fatigue driving detection system comprises a 3D structured light module, a central processing unit module and an alarm system; the 3D structured light module is installed in the front of a vehicle driver seat; the 3D structured light module is used for shooting face characteristics of a driver during driving,and transmitting data shot by the 3D structured light module to the central processing unit module; the central processor module is used for detecting, analyzing and judging whether the driver is in afatigue driving state or not by adopting a bagging algorithm, and sending a signal to the alarm system if the driver is detected to be in fatigue driving; and the alarm system sends out a warning signal to inform the driver not to be in fatigue driving. According to the system, the identification precision is high; the error is small; the driving state of the driver can be judged effectively andtimely; measures are correspondingly adopted when the driver is in fatigue driving; and dangerous driving of the driver is avoided by means of reminding the driver and the like.

The invention relates to an EV charging station position generation method and device based on a random particle swarm algorithm. The method comprises the steps of S1, collecting geographic information data; S2, based on the geographic information data, constructing an objective function and constraint conditions of an optimization problem according to the power distribution network factors, the traffic factors and the social factors; and S3, solving an optimization problem by combining a PSO algorithm and a Bagging algorithm, and generating an optimal charging station position. Compared withthe prior art, the method has the advantages that the charging station planning model taking cost and the planning effect as targets is adopted, and the PSO and Bagging algorithms are combined, so that the optimal site selection scheme of the charging station is realized, the user experience of the obtained charging station can be improved, and the energy consumption of idle driving is reduced.

The invention discloses a full-waveform LiDAR point cloud classification method based on multi-wavelet support vector machine WSVM integration. The method comprises the following steps: firstly, decomposing full-waveform LiDAR data, extracting geometric and waveform characteristics of point cloud, constructing a plurality of wavelet support vector machine (WSVM) classifiers by utilizing differentwavelet kernel functions, and constructing an integrated system by taking the WSVM classifiers as base classifiers; performing parameter optimization in each WSVM classifier by adopting a particle swarm optimization PSO algorithm, and integrating the WSVM classifiers by adopting a bagging algorithm; and finally, adopting a wavelet support vector machine (WSVM) classifier integrated system to classify point clouds decomposed from the original full-waveform LiDAR data, and outputting a result. According to the invention, a good point cloud classification result can be obtained.