164 results about "Fuzzy clusters" patented technology

The invention provides an intelligent fusion identification network state prediction and congestion control system. The invention discloses a method and a system for realizing network state predictionand data packet queue congestion control on a programmable data plane by combining a machine learning model method based on P4. The method comprises the following steps: collecting network state feature information in real time through an in-band network telemetry (INT) technology; and achieving network state characteristic value prediction by adopting an 'LSTM-fuzzy clustering' model method combining a long short term memory (LSTM) neural network model and a fuzzy clustering algorithm, and achieving the prediction of the network state characteristic value by adopting a fuzzy clustering algorithm. According to the obtained network state characteristic value, fuzzy clustering is carried out to obtain four network states; a normal state, a congestion early warning state, a continuous congestion state and a congestion alleviation state; corresponding strategies are formulated for different network states, the controller issues corresponding flow tables and formulates switch actions in different network states, and a comprehensive and dynamic queue feedback mechanism is provided to ensure that data packet queuing delay is as small as possible and achieve the effect of congestion control.

The invention discloses a fuzzy clustering analysis method-based method for determining a critical rainfall threshold of a landslide. The method comprises the steps of building an algorithm empirical model, determining the critical rainfall threshold of the landslide, and performing fuzzy clustering factor selection through a Delphi method; based on this, performing fuzzy clustering analysis of the critical rainfall threshold of the landslide, creating a data matrix and a standard data matrix, and establishing a fuzzy similar matrix; and finally performing fuzzy clustering and determining the critical rainfall threshold. According to the fuzzy clustering analysis method-based method for determining the critical rainfall threshold of the landslide, the rainfall threshold is determined by fully considering influence factors of regions through strict mathematic algorithm derivation, statistics technologies and physical means. The method is high in accuracy; the requirements of current landslide forecasting and warning can be met; the threshold accuracy is greatly improved; and reasonable determination of the critical rainfall threshold plays an important role for guaranteeing the landslide forecasting and warning accuracy.

The invention discloses a short-term photovoltaic power generation prediction method considering the correlation degree of weather and meteorological factors. The method comprises the following steps:1, rejecting bad data through an iForest algorithm; 2, respectively calculating Pearson correlation coefficients R of the photovoltaic power generation power and the five meteorological factors underfour weather types, and normalizing the Pearson correlation coefficients R; 3, performing fuzzy clustering on the five meteorological factors of the to-be-measured day, and obtaining a correlation coefficient of the historical day and the to-be-measured day; 4, a correlation coefficient normalization value is introduced, and the correlation degree between a historical day and a to-be-measured dayis solved; and 5, taking the historical day with high correlation degree as historical data, inputting the historical data and the meteorological factors of the day to be measured into the improved ACO-BP neural network, and finally obtaining a predicted value of the solar photovoltaic power generation to be measured; and 6, determining a neural network correlation coefficient, and performing simulation. The method aims to improve the photovoltaic power generation prediction precision, improves the practicality of a prediction model, and plays a great role in the combination of scheduling andprediction.

The invention discloses a single-cell transcriptome sequencing data clustering recommendation method based on two-dimensional distribution structure judgment, which comprises the following steps of: acquiring a gene expression matrix obtained by single-cell transcriptome sequencing data of a plurality of cells, and after filtering and standardization processing, constructing a two-dimensional feature matrix and carrying out linear normalization; calculating an Euclidean distance between cells according to the normalized two-dimensional feature matrix, thereby establishing a cell minimum spanning tree; cutting the cell minimum spanning tree through a self-adaptive threshold, and determining a two-dimensional distribution structure of data according to the balance of clusters formed after cutting; and recommending and applying a hierarchical clustering algorithm for data with fuzzy inter-cluster boundaries and a continuous two-dimensional distribution structure, and recommending and applying a spectral clustering algorithm for data with obvious inter-cluster boundaries and a block two-dimensional distribution structure. According to the method, a method which is more suitable for a two-dimensional distribution structure of single cell transcriptome sequencing data in hierarchical clustering and spectral clustering can be recommended to serve as a downstream clustering analysis method, and the clustering accuracy is improved.

The invention discloses an equivalent range-of-motion (ROM) distance calculation method based on a dynamic-time-warping (DTW) cluster fuzzy clustering self-organizing-maps (SOM) neuron. A feature vector is established, the feature vector is processed by an initial-state Kalman filter algorithm to remove secondary Gaussian covariance white noises to obtain data, and then a sample sequence for a dynamic time warping algorithm is obtained; the sample sequence is transformed into X-axis and Y-axis amplitude values of an acceleration signal acceleration amplitude SVM and the SVM amplitude values are classified into different groups by using a fuzzy clustering neuron algorithm; the group values of the two axial directions of the SVM(x) and the SVM(y) are combined with group clustering centers and gravity centers corresponding to the group values relatively; and then an ROM equivalent value is obtained by combining the group clustering gravies and the centers and a distance is calculated. Therefore, spatial offset distances between different group sensors can be found out and the path time of the end-to-end posture sensor is reduced.

The invention discloses a household energy double-layer optimization method for realizing interaction between a power grid side and a user side. The method comprises the steps of: reading a daily loadcurve of a user, carrying out dimensionality reduction on high-dimensional load data based on a feature index, and obtaining a dimensionality reduction curve; calculating the membership degree of thedimension reduction curve based on an FCM fuzzy clustering algorithm, and completing the classification of the daily load curve; classifying resident users, and designing a power package for each type of resident users; simulating a day-ahead load curve, and performing real-time monitoring and load decomposition on the household appliances of the users by utilizing a non-intrusive optimization model; dynamically adjusting a peak-valley electric quantity coefficient, and switching on and switching off equipment in real time according to the dynamic priority of the household appliances. According to the household energy double-layer optimization method for realizing interaction between the power grid side and the user side of the invention, the power package considering the time-of-use electricity price and an incentive subsidy mechanism is implemented to the scheduling of the specific load equipment of the household users, and is used for improving the peak-valley difference of the resident load curve, improving the interaction degree between a power utilization end and a power supply end, and enabling the resident users to participate in the optimal scheduling of the power grid more actively.

The invention provides a thermal ablation area monitoring method and system based on radio frequency processing and fuzzy clustering. The thermal ablation area monitoring method based on radio frequency processing and fuzzy clustering comprises the following steps: S1, acquiring radio frequency data in a thermal ablation process, and processing the acquired radio frequency data; s2, related to fuzzy clustering; s3, performing iterative processing on fuzzy clustering according to the processed radio frequency data and set fuzzy clustering parameters to obtain a clustering result; and S4, displaying a thermal ablation area according to the obtained clustering result. On one hand, the difference between thermal ablation and normal tissue is maximized and highlighted from the perspective of signal processing; on the other hand, an unsupervised fuzzy clustering algorithm is used for automatically identifying the thermal ablation area, the scheme can be well compatible with a current ultrasonic system, and compared with other disclosed thermal ablation monitoring methods, the method has the advantages of being reliable in result, easy to operate and easy to implement.

The invention discloses a confidence rule automatic generation method based on fuzzy clustering, and belongs to the technical field of data mining. In order to solve a problem that time and labor areconsumed when a confidence rule base is established completely depending on artificial experience, on the basis of an extended confidence rule base method, firstly, fuzzy clustering is carried out onsample data, and a fuzzy clustering center of a sample and the sample with the clustering center as the circle center and the distance between the sample and the clustering center being a certain setvalue serve as basic data for generating an extended confidence rule base; and calculating a premise attribute weight and a rule weight according to the variable association relationship and the sample membership matrix. The objective of the invention is to solve the problems of difficult selection of sample data and unsuitability for large sample data modeling in the extended confidence rule basemethod. According to the method, the dependency degree of experience setting premise attribute weights and initial rule weights in the existing extended confidence rule base method on expert experience is reduced. By adopting the method, the confidence rule is extracted from the actual production data, and compared with manually set initial confidence rules, the generation efficiency and the reasoning precision of the confidence rule base are improved.

The invention relates to data mining in bioinformatics, and particularly to a method for identifying the cancer-related miRNA through a miRNA-gene regulation module. The method comprises the steps ofperforming difference comparison of gene expression data; processing the gene expression data and miRNA expression data; constructing the miRNA-gene interaction matrix; calculating a miRNA-gene correlation coefficient, obtaining a miRNA-gene correlation matrix, performing fuzzy clustering on the miRNA; constructing a combined miRNA-gene interaction matrix and a miRNA-gene correlation matrix, calculating absolute average correlation degree of the gene with each miRNA, adding the gene into the miRNA according to absolute average correlation degree for constructing the miRNA-gene regulating module; calculating the correlation degree of the miRNA in each module, and ordering according to the correlation degree. The main process is presented in a graph 1. The method can be used for acquiring the cancer-related miRNA for searching the function and the mechanism in a cancer development and generating process, screening the miRNA biological mark in cancer early-period diagnosis, and acquiringtargets in targeted treatment of the cancer.

The invention discloses an urban expressway traffic state recognition method based on machine learning. The method comprises the following steps of S1, obtaining flow, speed and occupancy parameter data influencing an expressway traffic state; S2, expressing a group of traffic flow parameters including speed, flow and occupancy on a space coordinate axis, randomly selecting a plurality of points as initial clustering centers, dividing a data set into four types, and randomly selecting four points as four types of initial clustering centers; S4, generating a plurality of chromosomes according to a set population size, performing fitness evaluation on each chromosome, constructing a fuzzy clustering traffic state division model combined with a genetic algorithm by adopting a strategy of clustering first and then classifying, clustering a large amount of data, and performing traffic state classification by using an SVM (Support Vector Machine) to obtain a classification result; therefore,classification boundaries can be found more easily, and data processing efficiency and classification accuracy are improved.