148 results about "Cluster labeling" patented technology

The invention discloses a power consumption behavior clustering analysis method and system based on big data, and relates to the field of data analysis, and the method comprises the steps: firstly fusing collected active power time sequence data, operation and maintenance logs and meteorological monitoring data, and achieving the construction of sub-sequence-level features through the alignment of a multi-scale sliding window and a time window; then, multi-source fusion feature embedding is extracted by using a pre-trained depth encoder, so that the characterization capability for complex behaviors and mutation events is remarkably improved; further, adaptive determination of the cluster number and the center is realized through a Dirichlet process driven Gaussian hybrid dynamic clustering model, soft distribution is performed on embedded vectors by adopting a probability attribution degree inference method, and different power consumption behavior modes are accurately identified; and finally, according to the clustering labels, typical power consumption behavior labels are automatically distributed to all the parks. Therefore, the refined and intelligent level of energy efficiency management of the industrial park is comprehensively improved.

The invention belongs to the field of data storage, and discloses a frequency access data storage migration method and device for big data, and the method comprises the steps: obtaining a data access log, carrying out the preprocessing of the data access log, and obtaining a time window statistical matrix; carrying out timeliness attenuation weighting calculation on a plurality of data access times in the time window statistical matrix to obtain a weighted access frequency vector; obtaining a multi-dimensional feature matrix based on the weighted access frequency vector; clustering the feature vector of each time window slice in the multi-dimensional feature matrix through a Gaussian mixture model to obtain a clustering tag vector; inputting the clustering label vector into an access frequency prediction model to obtain an access frequency prediction value of the data in a future preset time step; the migration decision matrix is determined based on the access frequency predicted value and the data storage cost parameter, data migration is carried out based on the migration decision matrix, the data migration complexity can be reduced, and the access speed and the storage cost are balanced.

The invention relates to the technical field of communication production management, and discloses a communication equipment production intelligent management system based on machine learning. The system comprises a production data acquisition module, a feature engineering construction module, a dynamic clustering analysis module, an anomaly detection engine module and a production decision optimization module. The production data acquisition module acquires multi-source sensor data in real time and converts the multi-source sensor data into a standardized sequence with a unified timestamp; the feature engineering module extracts a time domain statistical feature, a frequency domain energy feature and an equipment state association feature to generate a high-dimensional feature vector set; the dynamic clustering module adopts an incremental algorithm to divide clusters online; the anomaly detection module establishes a multi-level Gaussian mixture model based on a clustering label, and quantifies an anomaly probability through a mahalanobis distance; and the production decision module integrates the results to generate an equipment maintenance priority sequence and a production takt adjustment instruction. According to the system, intelligent monitoring and dynamic optimization of the whole production process of the communication equipment are realized, and the real-time change requirement of a complex production environment is met.

The invention relates to the technical field of computer vision, in particular to a three-dimensional point cloud semantic segmentation method based on object clustering information. The method comprises the following steps: 1, acquiring three-dimensional point cloud position information of a vehicle surface, 2, processing point cloud data of a laser radar acquired in the step 1, and 3, clustering historical frame point cloud foreground semantic tags obtained through network inference in the step 2, the method comprises the following steps of 1, obtaining a priori clustering label, 4, sending a point subjected to downsampling in the step 2 and KNN nearest point information of the point into a network for feature extraction, 5, carrying out clustering feature fusion on the priori clustering label obtained in the step 3 and the point feature obtained in the step 4, and 6, predicting the point feature obtained in the step 4 and the clustering feature obtained in the step 5. And step 7, performing voting based on voxels by integrating the historical frame point cloud and the semantic tags of the current frame point cloud obtained by the six modules.

The invention provides a rock failure short-term and temporary early warning method and device based on acoustic emission clustering analysis, and relates to the technical field of rock mechanics and geotechnical engineering. The method comprises the following steps: acquiring acoustic emission characteristic parameters in a rock deformation and failure process; carrying out clustering analysis on the acoustic emission characteristic parameters by adopting a Kmeans + + algorithm to obtain clustering labels corresponding to the acoustic emission characteristic parameters; according to the clustering labels, a random forest algorithm is adopted, and importance scores of the acoustic emission characteristic parameters are calculated; constructing an early warning index set according to the acoustic emission characteristic parameters; an initial CNN-LSTM model is constructed; according to the early warning index set, establishing a sample data set of the rock failure conventional signal and the precursor signal; the initial CNN-LSTM model is trained according to the sample data set, and a trained CNN-LSTM model is obtained; acquiring a real-time acoustic emission signal; and inputting a real-time acoustic emission signal into the trained CNN-LSTM model, and carrying out early warning by identifying the category of the acoustic emission signal. According to the invention, early warning can be carried out in the rock complex damage process.

The invention relates to an electric power multi-source data slicing processing method and system based on reinforcement learning, and the method comprises the following steps: S1, obtaining electric power multi-source data, and carrying out the preprocessing of the electric power multi-source data, and obtaining a structured original data stream; s2, obtaining a multi-dimensional feature vector through feature extraction; s3, adopting a clustering algorithm to identify a behavior pattern to obtain a clustering label, calculating the distance from a sample to each clustering center, and when the distance is obviously higher than the average value plus threshold value beta times standard deviation of the cluster, additionally marking as abnormal; s4, intelligent slicing decision making is carried out based on reinforcement learning, and a slicing strategy, data and labels are obtained; s5, according to the obtained slicing strategy, data and label, through knowledge graph service rule verification, it is ensured that domain constraints are met; and S6, sending the verified strategy to distributed slice storage and management, and carrying out data slice storage and index construction according to the strategy. According to the method, the structured governance level and the high-value utilization capability of the power data assets are effectively improved.

The invention provides a generative adversarial network-based medical image transfer function automatic generation method, which comprises the following steps of: firstly, initializing a transfer function by using a Gaussian primitive, and carrying out multi-angle sampling on a rendering lens of volume data to obtain an optimal observation view angle of the volume data; generating rendering images of different transfer functions according to the optimal view angle and the transfer function by using a ray casting technology; secondly, selecting edge and gradient features as a clustering basis, and clustering rendered images of different transfer functions by using a k-means algorithm to obtain clustering tags; then, a CGAN-TF model is constructed, and the CGAN-TF model is trained by using a transfer function and a clustering label; and finally, selecting the generated transfer function by using a transfer function exploration component, and exploring and modifying according to requirements to obtain the transfer function with good performance and a corresponding visual rendering image. The method has the capability of generating the corresponding transfer function according to the label, and the stability is high.

The invention relates to a manufacturing system risk control knowledge matching method based on semantic embedding and clustering analysis, and the method comprises the following steps: collecting and preprocessing risk control text data: collecting unstructured text data of a manufacturing system history record, and obtaining preprocessed risk control text data, constructing a professional corpus for a discrete manufacturing scene; text semantic embedding generation; semantic clustering modeling: performing unsupervised clustering modeling on all semantic vectors, mining semantic association and potential structures between texts, obtaining semantic representations of risk control knowledge through a clustering algorithm, and assisting in generating clustering tags; and a risk knowledge matching mechanism.

The invention provides a public opinion video tag aggregation method and system based on artificial intelligence, and relates to the technical field of artificial intelligence. Comprising the following steps: acquiring pictures and text information in a short video, and performing semantic alignment; different large language models are adopted to generate preliminary labels for the pictures and the text information after semantic alignment; clustering the pictures and the texts after semantic alignment to obtain clusters; calculating the labeling probability of each primary label type in the current cluster by each large language model, and selecting the primary label with the highest probability sum as a clustering label of the current cluster; calculating the reliability weight of each large language model in the current cluster based on the clustering label of the current cluster; and based on the reliability weight, calculating the weighted support degree of all the large language models to different preliminary label types of each piece of data in the current cluster, calculating the weighted label of the current data, and further determining a final label. According to the method, the condition of few labels or no labels can be effectively processed, and the manual workload is greatly reduced.

The invention discloses a method for identifying traditional Chinese medicinal materials by combining infrared spectroscopy with clustering analysis, which comprises the following steps of: acquiring original infrared spectral data, averaging the original infrared spectral data to obtain single-sample original spectral data, constructing a sample graph and a wavelength graph based on standardized spectral characteristics, and fusing the sample graph and the wavelength graph to obtain the single-sample original spectral data. Obtaining fusion image data, inputting the fusion image data into a pre-trained image neural network model, extracting a low-dimensional feature vector of a target sample through forward propagation, and inputting the low-dimensional feature vector into a pre-trained dynamic cluster diffusion module to obtain a clustering label and distance data; and determining and outputting the quality grade of the rhizome traditional Chinese medicinal material sample based on the clustering label and the distance data. Therefore, interference can be effectively reduced, key features can be extracted, associated features of fused graph data are extracted in combination with a graph neural network, and the quality grade of traditional Chinese medicinal materials can be accurately determined in combination with clustering analysis of a dynamic cluster diffusion module.

The invention relates to a multi-granularity interest point recommendation method based on a geography-time mixed tree structure. The method comprises the following steps: collecting sign-in data of each user; clustering the geographic positions of the sign-in points to obtain a clustering result containing a plurality of geographic clusters, dividing each piece of sign-in data into each geographic cluster, and distributing corresponding geographic cluster labels; segmenting time, counting geographic clustering labels of all sign-in points in all time periods to obtain a track-level geographic hierarchical structure, and organizing the track-level geographic hierarchical structure into a tree structure to obtain a geographic tree and a time tree; a multi-level recurrent neural network is adopted to establish a recommendation model, geographic features and time features are obtained based on the geographic tree and the time tree, the input of the recommendation model is the geographic features and the time features, and the output of the recommendation model is a prediction result of interest point recommendation; inputting the spatio-temporal information of the current user into the trained recommendation model, and outputting the predicted next potential interest point by the recommendation model; and a more accurate interest point recommendation result is obtained.

The invention provides a multi-modal false news detection method based on unsupervised clustering and frequency domain information, and relates to the technical field of image text data. The method comprises the following steps: based on multi-modal sample data, respectively extracting text features, visual features and image text features, obtaining text feature clustering labels according to text feature global semantic correlation, inputting the text features with the labels into an unsupervised clustering learning network, and obtaining an unsupervised clustering learning result; semantic consistency is enhanced through a bidirectional gating loop unit and a multi-head attention mechanism, and enhanced text features are obtained; extracting frequency domain features based on the visual features, and fusing the frequency domain features with the spatial domain features to obtain visual joint features; and after image text features are also enhanced by the unsupervised clustering learning network, the image text features are fused with text and visual features through an attention fusion module by taking the image text features as a bridge to obtain multi-modal fusion features, and the multi-modal fusion features are input into a full connection layer to complete false news classification. According to the method, the accuracy of false news detection is effectively improved.

The invention provides a non-technical line loss state evaluation method based on multi-scale spatial-temporal feature fusion in a low-voltage distribution network environment, and relates to the technical field of electric power big data analysis and intelligent operation and maintenance of a distribution network. According to the invention, a fusion architecture based on a multi-scale time convolution network and a long short-term memory network is constructed; extracting multi-scale spatio-temporal characteristics of a user from instantaneous electricity utilization abrupt change to a periodic load rule through an MSTBlock unit; designing a cluster balance constraint mechanism to ensure that rare and key non-technical line loss abnormal early warning signals are not covered by mass normal power utilization data; according to the data scale, adaptively selecting a graph segmentation or spectral clustering integration strategy to output a clustering label, and mapping the clustering label into a user power consumption behavior evolution track; according to the method, the power utilization abnormal level can be identified from the original load signal with random fluctuation interference, and the troubleshooting priority is calculated in combination with the transformer area correlation analysis, so that the accuracy and interpretability of the non-technical line loss unsupervised evaluation decision of the power distribution network are remarkably improved.

The invention discloses a personalized teaching course recommendation method and system based on artificial intelligence, and relates to the technical field of artificial intelligence and education recommendation, and the method comprises the steps: collecting text data for preprocessing, and extracting a standard target set, a teaching target candidate set and a knowledge point candidate set; calculating cosine similarity and weight based on the standard target set and the teaching target candidate set, calculating knowledge point mastery degree and weight based on the knowledge point candidate set, and splicing the knowledge point mastery degree and weight to generate a learning feature vector; based on the learning feature vector, clustering is carried out by using a k-means + + algorithm, a clustering label and a clustering center set are output, after the clustering center is updated in combination with a Thompson Sampling algorithm and Monte Carlo, the posterior probability is recalculated, and a final recommendation result is output; the robustness and recommendation accuracy of personalized teaching course recommendation are effectively improved.

The invention discloses a charging pile charging power prediction method based on double-layer clustering. The method comprises the following steps: acquiring private charging pile order data; performing data cleaning and feature processing on the order data; a density clustering algorithm (DBSCAN) is adopted to cluster the charging features of the order data dimensions, and key features of the orders are extracted; clustering the charging behavior time sequence of the user by adopting a K-Shape clustering algorithm, extracting the charging behavior time sequence of the user, and standardizing the charging behavior time sequence; integrating the clustering results of the order and the user charging behavior time sequence to generate a comprehensive user-order behavior group, and obtaining a clustering label of global new-phase model input data; and taking the clustering labels as input data, and inputting the input data into a global new-phase model to predict the charging power of the private charging pile. The problem that power prediction is not accurate due to the fact that an existing power prediction method cannot cope with uncertainty caused by user behavior differences and personalized features in a private charging pile scene is solved.

The invention provides a k-plane clustering roof segmentation method based on multiple geometric consistency constraints. The k-plane clustering roof segmentation method comprises the steps of performing initialization preprocessing on input original building roof point cloud data; based on each point in the original building roof point cloud data and a candidate fitting plane cluster, establishing a clustering label optimization objective function fused with triple geometric constraints; iteratively updating the clustering label and the candidate fitting plane parameter of each point by adopting an alternating minimization optimization strategy, and solving a mixed integer non-convex optimization problem until the clustering label optimization objective function is converged; and based on the converged clustering tag optimization objective function, outputting a building roof point cloud instance segmentation result, the building roof point cloud instance segmentation result comprising the clustering tag of each point and the corresponding candidate fitting plane geometric parameters. According to the method, the segmentation precision and the calculation efficiency of the building roof point cloud instance are improved.

The invention discloses a single cell clustering statistics significance test method and system based on manifold learning, and the method comprises the following steps: obtaining original single cell data, carrying out the standardization processing of the original single cell data, and removing the deviation and noise in the data; performing dimension reduction on the standardized data by using a manifold learning method; dividing the dimension reduction data based on a pre-calculated clustering label, and performing hierarchical clustering on the divided data to obtain a hierarchical tree; hypothesis testing is carried out step by step from the bottom to the top of the hierarchical tree by adopting a Monte Carlo simulation method, and whether the clusters are combined or not is determined according to a hypothesis testing result; and finally, taking the combined hierarchical tree as a single cell clustering correction result. According to the method, the reliability and the accuracy of the clustering statistical test can be improved, and unnecessary group splitting and misjudgment can be avoided.

The invention provides an eye movement trajectory analysis method and system based on hybrid clustering and time constraint, and belongs to the technical field of computer vision. Calculating a time difference and a moving speed between continuous original eye movement data points; comparing the moving speed with a speed threshold value, and classifying the moving speed into candidate fixation points and glancing points; extracting spatial features and time features of the candidate fixation points, performing standardization processing, and performing weighted fusion to obtain spatial-temporal feature vectors; clustering the spatio-temporal feature vectors to obtain preliminary clustering labels of the candidate fixation points; performing time constraint processing on each cluster; calculating the duration of the clustering cluster after the time constraint processing, and generating a final clustering label of the candidate fixation point; and outputting a final classification label of each original eye movement data point in combination with the classification result of the glancing points.

The invention belongs to the technical field of analysis of load fluctuation behaviors, and particularly relates to an electrical load analysis method, system and device based on subspace clustering-CatBoost and a storage medium. Electrical load data are collected and subjected to standardization processing; clustering the standardized data through a subspace clustering algorithm to generate a clustering label; the clustering labels are classified through a CatBoost classifier, and the fluctuation behavior of the electrical load is analyzed; by performing standardization processing on the electric load data, the inconsistency of data scales is eliminated, the stability of data input is improved, different types of load modes are automatically identified through a subspace clustering technology, the classification of the load data is more accurate, meanwhile, the calculation efficiency of data processing is improved, and through a CatBoost classifier, the calculation efficiency of data processing is improved. Different types of load modes are accurately classified, the intelligent level of electricity consumption prediction is improved, and the interpretability of the model is enhanced.

The invention discloses an enterprise credit evaluation analysis method, and relates to the field of data processing, and the method comprises the steps: obtaining the credit data of a plurality of to-be-evaluated enterprises; performing feature extraction on the credit investigation data through an improved stack type self-encoding neural network model to obtain feature data; the improved stack type self-encoding neural network model comprises an encoder, a decoder, an attention layer and a prior rule layer; the attention layer adopts a Scaled Dot-Product Attention mechanism to learn a weight feature matrix Ha, and the priori rule layer sets feature constraints according to a priori rule; clustering the feature data to obtain a cluster to which the enterprise belongs; according to the feature value distribution of the enterprises in the clustering clusters, performing feature scoring by using a prior rule to obtain clustering labels of the clustering clusters; according to the clustering label and the weight feature matrix Ha, generating an enterprise credit investigation portrait, and according to the enterprise credit investigation portrait, carrying out credit investigation evaluation analysis; aiming at the low enterprise credit evaluation precision in the prior art, the reliability of the evaluation result is improved.

The invention discloses a scientific literature topic discovery method and device based on LLM. The method comprises the following steps: 1) acquiring text representation of each scientific literature sample and encoding the text representation by using a text encoder to obtain a document representation matrix corresponding to the scientific literature sample; 2) clustering the scientific literature samples to obtain clustering results of different themes; calculating an entropy value of each scientific literature sample, and selecting a high-uncertainty sample; 3) calculating semantic similarity between each high-uncertainty sample and other scientific literature samples, and constructing a plurality of triple tasks; fine-tuning the text encoder by using each triple task through a comparative learning method; 4) encoding the text representation of each scientific literature sample by using a text encoder to obtain a document representation matrix corresponding to the scientific literature sample; and 5) performing topic clustering on each scientific literature sample by using the document representation matrix of each scientific literature sample, and generating a clustering tag and a topic division result of each scientific literature sample.

The invention discloses a large-scale unstructured data joint processing method and system. The method comprises the steps that S1, large-scale unstructured data is collected, and a metadata feature library containing dynamic updating and index retrieval optimization is constructed; s2, constructing a federated index based on the dynamic metadata feature library, constructing a cross-domain joint index on the basis of the federated index, and performing layered federated architecture optimization to obtain a cross-domain joint optimization index; s3, constructing a global correlation matrix based on the cross-domain joint optimization index to perform cross-domain correlation modeling, and obtaining a global correlation function through cross-domain correlation modeling; and S4, carrying out joint optimization solution based on the global correlation function to realize unstructured data joint processing, uniformly optimizing tasks such as clustering, label prediction and correlation rule mining by fusing a processing target function, and enabling the tasks to cooperate with each other, so that the whole data processing process is more intelligent and efficient, and the accuracy of a data processing result is improved.

A system and method for determining a prioritized array of associated datasets. The system includes at least a processor and a memory communicatively connected to the at least a processor and contains instructions, wherein the at least a processor is configured to receive a plurality of datasets, apply a clustering module to the plurality of datasets, wherein the clustering module is configured to assign an individual dataset to an appropriate cluster, apply a classification module to the plurality of datasets, wherein the classification module is trained on cluster labels of one or more clusters and configured to predict labels for new individual datasets, and generate a prioritized array, wherein generating the prioritized array includes applying the plurality of instances of the plurality of datasets across one or more axes, wherein the one or more axes are derived from the clustering and classification of the plurality of datasets.

This invention provides a method and device for short-term early warning of rock failure based on acoustic emission clustering analysis, relating to the fields of rock mechanics and geotechnical engineering. The method includes: acquiring acoustic emission characteristic parameters during rock deformation and failure; performing clustering analysis on the acoustic emission characteristic parameters using the K-means++ algorithm to obtain cluster labels corresponding to the acoustic emission characteristic parameters; calculating the importance scores of the acoustic emission characteristic parameters based on the cluster labels using the random forest algorithm; constructing an early warning index set based on the acoustic emission characteristic parameters; constructing an initial CNN-LSTM model; establishing a sample dataset of conventional and precursor signals of rock failure based on the early warning index set; training the initial CNN-LSTM model using the sample dataset to obtain a trained CNN-LSTM model; acquiring real-time acoustic emission signals; and inputting the real-time acoustic emission signals into the trained CNN-LSTM model to provide early warning by identifying the acoustic emission signal category. This invention can provide early warning for complex rock failure processes.

The invention relates to the technical field of data processing and machine learning, and discloses a multi-modal clustering method, device and equipment, a storage medium and a product, and the method comprises the steps: determining the specific representation of original multi-modal data based on an auto-encoder and reconstruction loss, and carrying out the reconstruction of the original multi-modal data based on the specific representation; obtaining a clustering prototype through an adaptive average pooling operation and an expectation maximization algorithm, generating common characterization of the original multi-modal data according to a cross attention mechanism and the clustering prototype, carrying out comparative learning on the specific characterization and the common characterization to obtain aligned common characterization, and clustering the aligned common characterization to obtain a clustering result; and obtaining a clustering label of the original multi-modal data. According to the method, the unique representation is extracted through the self-encoder, the representation of the clustering prototype is dynamically adjusted by adopting the self-adaptive average pooling operation and the expectation maximization algorithm, and the common representation is generated based on the cross attention mechanism and the clustering prototype, so that the clustering label is obtained, the model data distinguishing capability is enhanced, and the accuracy of the clustering label is improved.

The embodiment of the application discloses a self-defined encryption protocol clustering method and device, the method comprises the following steps: obtaining a target encryption protocol to be identified, and extracting traffic data and metadata of the target encryption protocol; constructing a label matrix and a value matrix based on the traffic data; and fusing the label matrix and the value matrix by using a multi-mode matching algorithm to obtain a clustering label. In this way, the self-defined encryption protocol clustering method provided by the application combines double-matrix combination with a pattern matching algorithm to realize clustering of self-defined encryption protocols. The method solves the problem that mixed self-defined protocols cannot be quickly and accurately clustered, and provides support for subsequent protocol analysis by quickly and accurately realizing clustering of self-defined protocols.

The embodiment of the invention provides a multi-view clustering method and device, equipment, a storage medium and a program product, and relates to the field of financial science and technology. Obtaining a view data set corresponding to each view; based on the similarity between the feature elements in each view data set, constructing a similar matrix of each view data set; multiplying the similar matrixes of the plurality of view data sets to obtain a consistency matrix, and reconstructing the similar matrix of each view data set based on the consistency matrix to obtain a reconstructed similar matrix; and performing multi-view clustering based on the plurality of reconstructed similar matrixes to obtain a clustering result, the clustering result being used for indicating a clustering label corresponding to each sample. Through the above mode, the complementary information of the consistency matrix is utilized to maintain the stability of the clustering boundary, the misjudgment and missed judgment caused by the missing of the sample information in the view are reduced, and the accuracy of the multi-view clustering result is improved.

The present invention provides a method and system for automatically partitioning a non-stationary training image based on depth features, the method comprising: cutting and dividing a non-stationary training image into a plurality of training image sub-blocks; extracting the depth features of each training image sub-block based on a convolutional neural network model; performing cluster analysis on the depth features of all training image sub-blocks to obtain a clustering label of each training image sub-block; splicing each training image sub-block according to its position in the non-stationary training image based on the clustering label of each training image sub-block to obtain a partitioning result of the non-stationary training image; performing a stationary algorithm simulation based on the partitioning result to output a simulation implementation of the non-stationary training image. The present invention automatically partitions the non-stationary training image, and compared with the SIMPAT algorithm based on the non-stationary training image without partitioning, the multi-point geological statistical simulation results after the partitioning based on the present invention better reproduce the non-stationary features.

The invention discloses an industrial park weather typing method and device integrating a self-organizing mapping neural network and a k-means algorithm, and the method comprises the steps: firstly, constructing an initial data set through air quality monitoring data and ERA5 meteorological reanalysis data, converting the data into a numerical value type, and carrying out the standardization, so as to guarantee that the feature data has the distribution with the mean value being 0 and the standard deviation being 1; thirdly, initializing a self-organizing mapping (SOM) model, and performing dimension reduction on the data by setting parameters such as grid size, sigma and learning rate so as to obtain two-dimensional feature representation; then, initializing a k-means clustering model, setting a clustering number and a random seed, and performing clustering analysis on the feature data after dimension reduction by using a fitpredict () method so as to obtain a clustering label of each sample; in the evaluation and visualization stage, evaluation indexes such as SSE, MSE, RMSE, a contour coefficient and a Clinski-Harabasz index are calculated so as to evaluate the clustering effect and the separation degree. And finally, adding a clustering label into the original data, and visually analyzing the distribution of each feature under different clusters by using a violin chart, thereby providing a visual basis for further analysis. According to the method, through integration of the SOM and the k-means model, the precision of ozone concentration weather typing of the small-scale industrial park is remarkably improved, and the high-concentration ozone weather is divided more finely and more accurately.

The invention discloses a new energy unit icing shutdown prediction method and system based on a multi-feature interaction threshold. The method comprises the steps of collecting and preprocessing multi-dimensional data of a new energy unit; based on meteorological and geographic features, generating a corresponding clustering label for each new energy unit by using a clustering algorithm; constructing a full-connection deep neural network shutdown prediction model, taking the preprocessed data and the clustering labels as input features, and training the icing shutdown probability of a model output unit; for a single feature, a feature fixing strategy is adopted, and a single feature threshold interval is determined; key features are selected for double-feature interaction analysis, a three-dimensional decision boundary is constructed, an interaction effect is quantified, and a multi-feature interaction rule is extracted based on a decision tree algorithm; and constructing a comprehensive discrimination rule, setting risk preference parameters, carrying out adaptive threshold updating, and finally outputting a shutdown prediction result. The method can significantly improve the shutdown prediction precision, and is suitable for different types of new energy equipment such as wind power and photovoltaic equipment.