57results about How to "Good dimensionality reduction effect" patented technology

The invention particularly relates to a wireless fidelity (Wi-Fi) indoor positioning method, and aims at resolving the problems that in a traditional Wi-Fi indoor positioning method, a feature information position fingerprint map data base is too large, computation complexity in an on-line positioning phase matching process is high, instantaneity is poor, and the like. The method includes that when a point to be detected receives a wireless signal strength value sent by a wireless connection point, a support vector machine classifier is adopted to position the point to be detected to a corresponding ith subregion, and a position fingerprint map and a feature transformational matrix Ai of the subregion are obtained; and the feature transformational matrix Ai of the ith subregion is adopted to enable the wireless signal strength value of the point to be detected to achieve shiftdim, a d-dimension wireless signal strength value is obtained and matched with the subregion, the weight K-nearest neighbor node algorithm is adopted to forecast position coordinates of the point to be detected, and positioning results are output. The Wi-Fi indoor positioning method is applied to the communication field.

The invention discloses an advanced persistent threat (APT) attack detection method based on deep belief network-support vector data description. A deep belief network (DBN) is used for feature dimension-reduction and excellent feature vector extraction; and support vector data description (SVDD) is used for the data classification and detection. At a DBN training state, the feature dimension-reduction is performed by using the DBN model after obtaining a standard data set; a low-level restricted Boltzmann machine (RBM) receives simple representation transmitted from the low-level RBM by usingthe high-level RBM so as to learn more abstract and complex representation after performing the initial dimension-reduction, and back propagation of a back propagation (BP) neural network is used forrepeatedly adjusting a weight value until the data with excellent feature is extracted. The data processed by the DBN is divided into a training set and a testing set, and the data set is provided for the SVDD to perform training and identification detection, thereby obtaining the detection result. The attack detection method disclosed by the invention is suitable for the unsupervised attack datadetection with large data size and high-dimension feature, is fit for the APT attack detection and can obtain an excellent detection result.

The invention discloses a feature selecting method for pattern classification. The method includes on the basis of figuring out a unsupervised optimal discriminant vector by adopting fuzzy Fisher criterion as an objective function, figuring out each feature importance weight according to a value of each dimension of the vector, collating the features according to dimension of the weights, selecting subsets of the features through a given threshold, and accordingly realizing feature dimension reduction. According to the feature selecting method, sample classify information is not required to provide beforehand, the problem that feature selection is lack of separation measurement in a unsupervised mode is solved effectively, good dimension reduction performances are realized in UCI data set and fault diagnosis experiments, and the feature selecting method is highly practical.

The invention provides a software vulnerability automatic classification method based on a deep neural network. The method comprises the steps of S1, preprocessing the vulnerability information to form a word set list; S2, calculating the weight of each word in the sample vulnerability description information set by using a TF-IDF algorithm and an IG algorithm to obtain an important feature word set list; S3, generating the word vector space according to the important feature word set list, expressing each piece of vulnerability description information as an m-dimensional vector, wherein m isthe number of feature words in the important feature word set; S4, obtaining a software vulnerability classifier by using the DNN model; and S5, classifying the new vulnerability description information set. According to the method, an automatic deep neural network vulnerability classification model is constructed based on a TF-IDF algorithm and an IG algorithm; the dimension of the high-dimensional word vector space is reduced, the method can adapt to continuously updated software vulnerability data sets, the high dimension and sparsity of the word vector space are effectively processed, andthe better performance is shown in multi-dimensional evaluation indexes such as the accuracy rate, the recall rate and the precision.

A method for evaluating enterprise suppliers based on mixed weight kernel principal component analysis, characterized in that: the method includes the following steps: (1) determining the supplier evaluation criteria, and classifying the output and input indicators; (2) the step ( 1) Take the reciprocal of the obtained input indicators and normalize all indicators using the maximum and minimum method, so that the range is between [0, 1]; (3) Check the normalized data obtained in step (2) Matrix calculation, calculate the averaged kernel space sample kernel matrix and find its eigenvalues ​​and eigenvectors, normalize the eigenvectors, select the eigenvectors with a cumulative contribution rate greater than 80% to form an orthogonal basis matrix; (4) The samples obtained in step (2) are projected, the mixed weight vector is calculated, and the corresponding projection coordinates are weighted and summed; (5) The comprehensive indicators obtained in step (4) are sorted in reverse order. The application of the present invention is applied to enterprise supplier evaluation.

The invention relates to an on-line target tracking method based on probabilistic principal component analysis and compressed sensing. A compressed sensing theory is combined with a PPCA (probabilistic principal component analysis) theory to obtain a quite simple target characteristic subspace representation model, and the target subspace representation model is updated on line by the aid of an incremental PCA (principal component analysis) algorithm to enable a tracking algorithm to achieve good real-time performance. Meanwhile, the PPCA theory is applied to computation of visual similarity of a candidate target and the target subspace representation model to obtain a DFFS (distance from face space) and a DIFS (distributed inter-frame space) not just reconstruction errors in PCA, so that visual similarity is computed more accurately, and further accuracy of the tracking algorithm is guaranteed.

The invention discloses an extra-large basin hydropower station group optimal scheduling method based on a hybrid intelligent algorithm. Hydropower stations involved in calculation are selected and corresponding constraint conditions are set, an individual serial coding method is adopted to code individuals, and an initial population is generated; the fitness value of each individual is evaluated,an individual extremum is subjected to mutation operation after the individual extremum and a global extremum are updated, the positions of all individuals in the population are then updated, and a hybrid search strategy is then carried out on an individual in an external file set to improve the individual diversity; and finally, the above process is repeated until a terminating condition is met.The optimization performance is excellent, the robustness is strong, the convergence speed is fast, programming for realization is easy, and the problem of curse of dimensionality of the traditionalscheduling algorithm is avoided. Examples of Wujiang basin application show that the method of the invention can effectively improve the individual convergence speed and the population global search ability, a reasonable and effective hydropower station group scheduling operation mode can be acquired quickly, and the overall scheduling benefits of the extra-large basin hydropower station group scheduling are improved.

The invention relates to a face rapid recognition method based on t distribution in a complex environment. The method includes a sample training step and a face recognition step. The sample training step specifically includes: performing feature extraction on a training sample through a convolutional neural network to obtain a feature sample; and projecting the feature sample to a low-dimensional space through t distribution non-linear projection to obtain the training sample in the low-dimensional space and dimension reduction path parameters. The face recognition step specifically includes: performing feature extraction on a to-be-recognized image through the convolutional neural network to obtain a feature sample of the to-be-recognized image; projecting the feature sample r of the to-be-recognized image to the low-dimensional space according to the dimension reduction parameters to obtain a test sample s in the low-dimensional space; and performing classification recognition on the training sample (y1, y2, ..., yn) and the test sample s in the low-dimensional space through a classifier to obtain a recognition result. Compared with the prior art, the method is advantaged by high recognition efficiency, more excellent recognition performance, and high flexibility.

The invention discloses a hyperspectral remote sensing image dimensionality reduction method based on conformal geometric algebra. The method includes the steps of firstly, collecting hyperspectral data, preprocessing the same, and the like; secondly, performing information expression on the hyperspectral data under a conformal geometric algebra space, and building the mapping relations among data of different spaces; thirdly, building a hyperspectral image feature distance operation operator; fourthly, building the expression method of distance measurement on the basis of the conformal geometric algebra; fifthly, calculating the distances of different wave bands, and k adjacent wave bands of each wave band; sixthly, using a Floyd shortest path calculation algorithm to calculate the shortest distance among the wave bands, and using the shortest distance as the matrix for dimensionality reduction; seventhly, using a PCA algorithm to calculate b feature values of the distance matrix, using the b feature values as the mapping coordinate systems, and taking the wave band data described by the coordinate systems as the to-be-selected wave band data. By the method, the hyperspectral remote sensing image feature extracting effect can be increased, and data information loss caused by the existing hyperspectral image data dimensionality reduction methods can be reduced.

The invention discloses an agency optimization dimension reduction method for combined scheduling of reservoir groups of a large-scale hydropower station and belongs to the technical field of optimization scheduling of hydropower systems. The method comprises the following steps: selecting a hydropower station, setting related constraints and parameters, and calculating an initial scheduling process O and a search step length h; according h and O, generating a sample set S1 within a neighborhood domain range, calculating real target functions of samples in S1, inputting sample points in S1 andthe target functions into a neural network, and carrying out fitting so as to obtain a corresponding agency optimization model N; generating a sample set S2 of a certain scale, inputting sample points in S2 into N so as to obtain possible target functions, screening an optimal sample set S3 according to the target functions, calculating real target functions of sample points in S3, finding an improvement solution with an optimal target function in S3, if the improvement solution is prior to an initial solution, updating the initial solution, and carrying out iterative calculation, or else judging whether a step length meets precise requirements or not, if the step length meets the precise requirements, outputting the optimal solution, or else updating and replenishing the initial solution, and carrying out iteration. The method disclosed by the invention is small in parameter calculation quantity, high in search precision, short in optimization time and large in dissolution scale.

The invention discloses a two-stage direct search dimension reduction method for optimal dispatching of cascade hydropower stations, belonging to the technical field of efficient utilization of waterresources and optimal dispatching of hydropower system. Includes such steps as decomposing multi-stage reservoir group dispatching problem into several two-stage sub-problems after initial dispatchingprocess is given and maximum search step length is calculated; Then the direct search strategy is used to solve each sub-problem. Finally, the global optimal solution is approximated successively byiterative optimization. The practice result of the cascade hydropower station group joint operation fully verifies the effectiveness of the method of the invention. Compared with the traditional stepwise optimization algorithm, The invention adopts the direct search strategy to replace the enumeration calculation operation in the sub-problem, reduces the calculation complexity from the exponentialgrowth to the polynomial growth, remarkably reduces the calculation time and occupies the memory, greatly improves the execution efficiency and the solution scale, and is more suitable for the large-scale complex hydropower system optimization dispatching problem.

The invention relates to a flight emergency prediction method and apparatus based on an LSTM neural network. The method comprises the steps of obtaining sample data of historical accident flight records; preprocessing the sample data, then performing dimension reduction on the sample data by utilizing an MDDM algorithm, and mapping the sample data into a multi-dimensional vector; constructing a prediction data set by taking the multi-dimensional vector as a sample; dividing the prediction data set into a training set, a verification set and a test set, and training an LSTM neural network until the error of the LSTM neural network is lower than a threshold value and tends to be stable; and inputting the current flight data into the trained LSTM neural network to obtain the flight emergency probability. According to the invention, related data of historical flight emergencies are subjected to preprocessing, dimensionality reduction and feature extraction and are used as samples to train the LSTM neural network, so that various data in the flight process are automatically and comprehensively monitored, early warning is given out for the flight emergencies in time, and safe execution of flight tasks is guaranteed.

The invention discloses a hyperspectral remote sensing image waveband selection algorithm based on improving firefly algorithm, and object functions in the FA algorithm are improved. Optimization and improvement of waveband selection is characterized by carrying out random initialization on waveband index position, position matrix size being s=n*b (n being known parameters, b being user-input waveband selection number); selecting different spectrum type distance function as the object function, substituting the obtained initial position matrix into the object function for calculation, and obtaining a group of one-dimensional array corresponding to the fluorescence brightness values of fireflies; carrying out ranking (disadvantaged point approaching to advantaged point) according to the advantages and disadvantages of the brightness values, that is, the value of the object function value; updating the waveband which is subjected to feature selection, that is, the position information of the fireflies after movement; and recording the waveband selection results when according with maximum iterations or searching precision. According to the hyperspectral remote sensing image waveband selection method based on firefly optimization, the problems that a conventional hyperspectral remote sensing image waveband selection algorithm is not high in precision and time-consuming can be solved; and the method has the advantages of being good in waveband selection effect, and wide in adaptation and the like.

The invention discloses a re-admission risk prediction method based on an adaptive ensemble learning model. The method comprises the steps of collecting basic information and clinical diagnosis and treatment information of a patient, and constructing a clinical high-dimensional feature matrix and a re-admission label; sequentially performing data preprocessing and KPCA dimension reduction on the clinical high-dimensional feature matrix to obtain a dimension reduction feature set; and constructing an adaptive ensemble learning model, training the adaptive ensemble learning model according to the dimension reduction feature set and the re-admission label, and inputting the dimension reduction feature set of the patient to be predicted into the trained adaptive ensemble learning model to obtain a re-admission risk prediction result of the patient. According to the re-admission risk prediction method based on the self-adaptive ensemble learning model, the re-admission risk of the patient is accurately predicted through the ensemble learning model, a doctor is assisted in taking intervention measures on a high-risk patient in advance, the disease burden of the patient can be reduced, the economic burden of the patient is reduced, the hospital re-admission rate can be reduced, and the medical service quality can be improved.

The invention discloses a birdsong automatic identification method based on feature fusion, and belongs to the technical field of ecological monitoring and acoustic signal identification. The method comprises the following steps: firstly, collecting different birdsongs, each birdsong including a plurality of birdsong segment; then, extracting frequency domain features of each birdsong segment; then, extracting image domain features of each birdsong segment; then, splicing the frequency domain features and the image domain features extracted from each birdsong segment correspondingly in a one-to-one manner, to form a high-dimensional feature set of multi-feature elements; secondly, selecting features of the high-dimensional feature set of multi-feature elements, to obtain a low-dimensionalfeature set of multi-feature elements; finally, identifying bird species by combining the low-dimensional feature set with a classifier. The method effectively overcomes shortcomings that a single feature is difficult to describe birdsong audio characteristics, the method has strong anti-background noise interference ability, low calculation complexity, short running time, and reliable data and iseasy to implement.

The invention discloses a hyperspectral image characteristic extraction algorithm based on manifold learning linearization and belongs to the technical field of hyperspectral image data processing and application. The shortcoming that a manifold learning algorithm has no generalization ability is overcome through the improved manifold learning linearization algorithm. The method comprises the steps that first, a preliminary dimensionality reduction result and a Laplacian matrix are computed; second, a matrix equation set constant term matrix and a coefficient matrix are established; third, a characteristic converting matrix is computed; and fourth, a final dimensionality reduction result is computed according to the characteristic converting matrix. The shortcoming that the global linear mapping hypothesis in LPP, NPE and LLTSA linearization manifold learning algorithms is invalid most of the time is overcome, a penalty term which deviates from an original manifold learning algorithm result is added into an original cost function, a bound term in an original target function is removed, and solving of the optimum characteristic transition matrix is converted into solving of a matrix equation set. The algorithm is suitable for hyperspectral image characteristic extraction.

The invention discloses an apple tree canopy scale nitrogen content diagnosis method. The method comprises the following steps: determining the reflectivity of an apple tree canopy scale by adopting ahyper-spectrometer; selecting an apple tree canopy leaf sample; determining the total nitrogen content of the apple tree canopy leaf sample; identifying abnormal samples in a reflectivity data set and a total nitrogen content data set by utilizing a Monte Carlo secondary detection method; preprocessing the reflectivity data set; aiming at the obtained reflectivity data sets of 1326 wave bands, extracting variables based on a potential variable method in a partial least squares regression analysis process; establishing a model for the extracted variables by adopting an extreme learning machine; and predicting the nitrogen content of the apple tree canopy leaves by utilizing the established model. According to the method, canopy scale spectral information is collected for nitrogen conditiondiagnosis, errors caused by sample randomness due to existing leaf scale diagnosis are effectively avoided, and the diagnosis result can measure the overall canopy nitrogen condition of the apple tree.

The invention discloses a low-voltage power distribution network three-phase load flow calculation method based on a phase sequence mixing method. The method comprises the steps of enabling a single-phase node to be equivalent to a three-phase node in a phase component model to form a three-phase symmetrical low-voltage distribution network; secondly, converting phase component model parameters ofthe power distribution network into sequence component model parameters, then performing load flow calculation by adopting a forward-derivation descendant method, converting an obtained load flow result into a phase component model, and finally solving single-phase node load flow through a previous equivalent superior three-phase node load flow result and parameters. According to the method, a phase component method is adopted for an asymmetric part, and a sequence component method is adopted for a symmetric part, so that the low-voltage power distribution network can also apply the symmetriccomponent method, the number of nodes, which participate in load flow calculation, of the system is reduced in the solving process, and the larger the number of single-phase nodes is, the better thedimension reduction effect is, and the higher the load flow calculation speed is.

The invention discloses an image processing and classifying method and system. The image processing method comprises the following steps: cutting main features of each sample image to obtain a plurality of local images containing the main features, and recording the local images as sub-images; combining the gray value vectors of the sub-pictures together to form a sample data matrix; constructinga weight coefficient representing the similarity degree between the pictures; determining a similarity matrix according to the weight coefficient; calculating a Laplace matrix; taking the minimum lossfunction as a target, and determining an optimal projection matrix according to the Laplace matrix and the sample data matrix; judging whether the loss function converges or not; and if not, updatingthe weight coefficient, and skipping to the step of determining the similar matrix, and if so, performing dimensionality reduction on the to-be-processed image by adopting the optimal projection matrix corresponding to the convergence of the loss function. According to the image processing method, the manifold structure embedded in the data can be reserved, and the image recognition method basedon the image processing method has the advantages of saving computing power and being high in recognition accuracy.

The invention relates to a greedy strategy based flight path planning method for an unmanned aerial vehicle, and in particular to the field of navigation calculation. The method comprises the following steps: S1, acquiring an original data set, and preprocessing the original data set by data set visualization and data dimensionality reduction to obtain a dimensionality reduction data set, a starting point A, a first destination point B and a level metric d; S2, starting from the starting point A, traversing sequentially along a spatial point set X = {x1, x2,..., xt, xt+1,..., xn} sorted by the levelness d; and S3, finding the farthest point xt which can be accessed by the starting point A, and sequentially constructing a set X'= {x1, x2,..., xt} that the starting point A can access the correction point under the condition that the constraint condition is met. According to the scheme, the technical problem of how to quickly plan the flight path under the constraint of system positioning accuracy is solved, and the method is suitable for flight path planning of aircrafts.