195 results about "Data dimensionality reduction" patented technology

The present application relates to the technical field of financial risk management, and in particular to a big data-based financial risk assessment and control system. By means of acquiring and preprocessing multi-source financial data of an enterprise, comprising data cleaning, data fusion, and data dimensionality reduction, the system generates standardized financial data. On the basis of the standardized financial data, constructing a multi-dimensional risk assessment model, comprising a market risk assessment model, a credit risk assessment model, an operation risk assessment model and a compliance risk assessment model. Using the multi-dimensional risk assessment model to perform risk assessment, generating a risk assessment report, and according to the report, generating a financial risk management suggestion, thereby implementing real-time alert and dynamic adjustment. The present invention improves the comprehensiveness and accuracy of financial risk assessment, achieves real-time monitoring and dynamic adjustment of financial risks, and improves the integration and reliability of the system.

The invention belongs to the technical field of industrial flue gas purification, and discloses an ammonia desulfurization optimization control system based on a machine learning algorithm, which comprises a feature extraction module, a working condition clustering module, a dynamic optimization module and a self-adaptive feedback module, the feature extraction module is used for collecting multi-dimensional operation parameters in a coal burning process, performing data dimension reduction through a PCA algorithm, and extracting key working condition features; the working condition clustering module identifies different operation working condition modes; the dynamic optimization module can construct a multi-modal optimization neural network based on different operation condition modes, and generates optimal control parameters in real time. Through PCA dimension reduction processing of the feature extraction module and in combination with a working condition clustering algorithm, automatic mode recognition and strategy switching under complex working conditions are achieved, fluctuation of desulfurization efficiency is reduced, and meanwhile the problems that the ammonia water adding amount depends on experience setting, raw material waste and secondary pollution are likely to be caused, and the operation and maintenance cost is large are solved.

The invention discloses a tractor transportation operation working condition construction method based on improved particle swarm optimization and KMeans fusion, and relates to the technical field of agricultural machinery working condition analysis. The method comprises the following steps: acquiring original data of tractor transportation operation through a plurality of data acquisition modes, and carrying out preprocessing and three-stage screening to obtain an effective kinematics fragment; selecting multi-dimensional characteristic parameters to construct a characteristic matrix, and performing data dimension reduction by adopting principal component analysis; optimizing a KMeans clustering initial center by using an improved particle swarm optimization (IPSO) algorithm which introduces a dynamic inertia weight and a Gaussian mutation strategy, and performing clustering analysis on the feature space after dimension reduction; and selecting representative fragments based on feature similarity, and synthesizing a standardized working condition curve by taking the sum of average relative errors of all feature dimensions as a target function. According to the method, the problems that a traditional clustering algorithm is prone to falling into local optimum and the working condition construction accuracy is insufficient are solved, the constructed working condition can truly and comprehensively reflect the actual transportation operation characteristics of the tractor, and a reliable basis is provided for tractor power system optimization, operation efficiency improvement and energy consumption reduction.

The invention discloses a risk early warning method, training method and device of a risk identification model. The method comprises the steps that N kinds of performance index data of an information system in a target time period are collected, and a target index data set is obtained; the target index data set is input into a risk recognition model, a risk recognition result is output, the risk recognition model is obtained after an initial recognition model is trained based on a target training sample, and the target training sample comprises data obtained by conducting dimension raising on historical index data in a historical index data set and data obtained by conducting dimension raising on the historical index data in the historical index data set; the historical index data comprises N kinds of performance index data of the information system in a historical time period; and generating early warning prompt information under the condition that the risk identification result indicates that the information system has the risk. According to the method and the device, the problem that a risk prediction result of an information system is inaccurate due to data dimension reduction when a machine learning algorithm in related technologies processes multi-dimensional data is solved.

The invention discloses a crop ralstonia solanacearum disease prediction method based on an integrated learning model, and the method comprises the following steps: S1, collecting a published 16s rRNA gene sequence related to solanaceae crop bacterial wilt, and carrying out the preprocessing of original sequencing data based on an EasyAmplicon standardized process; s2, performing data dimension reduction by using a principal component analysis algorithm, and retaining 95% of variance; s3, performing hyper-parameter search based on 5-fold cross validation and grid search on the Light GBM model, the CatBoost model and the XGBoost model respectively, and selecting three groups of optimal hyper-parameters of each model; s4, constructing a model according to three groups of optimal hyper-parameters of each selected model, performing prediction, analyzing model result difference based on a Pearson correlation coefficient, and retaining Pearson correlation coefficient mean < lt > with other eight model prediction values; a model of 0.8; and S5, inputting the screened model prediction result into a second-layer element learner RF for integrated learning to obtain a final prediction result.

The invention belongs to the technical field of battery health degree evaluation, and particularly relates to a battery health degree evaluation method and system, electronic equipment and a storage medium. Based on the original sample data of the battery health degree evaluation indexes, establishing a standardized decision matrix to eliminate the dimensional difference between the indexes; constructing a correlation matrix according to the matrix, and evaluating index correlation; extracting common factors of a correlation matrix, calculating a core quantitative index, determining a main factor by taking standard reaching of an accumulated variance contribution rate as a standard, realizing data dimension reduction and retaining original index core information; and calculating a main factor score in combination with the original data and the rotated factor matrix, performing weighting according to a variance contribution rate, and performing weighted summation to obtain a health degree comprehensive score. And the health difference of the batteries can be judged in an auxiliary manner through clustering analysis, and data support is provided for reasonable configuration and fine management of the batteries. According to the method, dimension reduction weighting is carried out on various related indexes through a factor analysis method, a result is objectively deduced based on data features, subjective influences are reduced, and evaluation scientificity and reliability are improved.

The invention relates to the technical field of foreign matter recognition, in particular to a foreign matter early warning method and device, electronic equipment and a storage medium. Carrying out dimensionality reduction on the plurality of first spectral vectors through a dimensionality reduction conversion matrix to obtain a plurality of second spectral vectors; classifying each second spectrum vector according to the relationship between the second spectrum vectors and a plurality of spectrum classes to obtain a plurality of spectrum class identifiers; and finally, selecting a plurality of target identifiers from the plurality of spectrum identifiers, determining a first foreign matter probability according to the plurality of target identifiers, the perimeter area ratio of the first region and a conditional probability equation, and performing early warning according to the first foreign matter probability. According to the invention, based on the visual identification area, the spectral data is extracted, dimensionality reduction and classification are carried out on the spectral data, the foreign matter risk is determined based on data dimensionality reduction, classification and statistics, the accuracy of foreign matter early warning is improved, and the probability of false alarm when the foreign matter is found is reduced.

The invention discloses a composite material reflectivity spectral information classification method and system based on a PCA-SVM algorithm. The method comprises the following steps: collecting reflectivity spectral data of a composite material sample; preprocessing data to eliminate measurement deviation and unify numerical scale; carrying out dimensionality reduction on the preprocessed high-dimensional spectral data through principal component analysis, and extracting feature components retaining main variance information; based on dimension reduction features, a support vector machine is adopted to construct a multi-label classification model according to a'one-to-other 'strategy; predicting the test sample, and generating a multi-label classification result through probability output and threshold processing; and analyzing the classification performance by using the multi-label evaluation index. The data processing module of the system executes preprocessing, dimension reduction, modeling, prediction and evaluation operations. The method is suitable for lossless identification of multi-component composite samples, both interpretability and identification precision are considered, the performance bottleneck of a traditional method under the conditions of feature overlapping, insufficient samples and the like is effectively overcome, and efficient and accurate classification of the composite materials is achieved.

The invention discloses a battery cell abnormity determination method and system based on BMS data driving, and the method comprises the steps: collecting multi-dimensional battery cell operation data through employing a plurality of sensors and a BMS, and carrying out the preprocessing of the collected data, and forming a basic data set; performing data dimension reduction and feature extraction on the obtained basic data set, comparing an extracted feature value with a dynamic adaptive threshold obtained based on a dynamic threshold adjustment mechanism to obtain a cell health state evaluation result, generating an abnormal early warning signal according to the cell health state evaluation result, and positioning a potential fault point; and performing exception classification on the exception early warning signals, generating maintenance suggestions, generating exception records and storing the exception records in a cloud database. According to the invention, accurate identification and rapid response to the abnormity of the battery cell are realized.

The invention discloses a small sample non-uniform multi-mode clutter modeling and partition suppression method, which is applied to the technical field of radars, performs multi-mode clutter modeling aiming at a complex environment, partitions the non-uniform clutter environment based on the difference of non-uniform clutter covariance matrix geometric matching, and improves the robustness of the multi-mode clutter modeling. Then, the problem that the performance of a traditional clutter suppression method is reduced under the small sample condition is solved through a dimensionality reduction self-adaptive filtering method in different areas; according to the method, firstly, multi-mode clutter echoes are generated for a complex environment, different-mode clutter region boundaries are judged through the difference of different clutter covariance matrix geometric matching, clutters are partitioned, and then the degree of freedom is reduced through data dimension reduction in different regions, and filter weight vectors are designed to achieve self-adaptive filtering; according to the method, effective partitioning and suppression of multimode clutters can be realized under the conditions that the number of samples is less than the degree of freedom of a radar system and the clutter environment is not uniform.

The invention discloses a laser cutting end surface quality control method, which comprises the following steps: S1, data acquisition: acquiring the thickness, material spectrum, surface roughness, laser power and auxiliary gas pressure data of a workpiece by using a multi-mode sensing module, performing dimension reduction on hyperspectral data through a principal component analysis algorithm, extracting a material feature vector, and calculating the material feature vector; and generating an initial process parameter request in combination with the thickness information. Through a multi-parameter dynamic collaborative optimization mechanism, a coupling rule among parameters in the laser cutting process is deeply excavated, parameters such as laser power, cutting speed, focus position and auxiliary gas pressure are brought into a unified regulation and control system, and through a parameter coupling relation model and a multivariable collaborative regulation strategy, compared with a traditional single-parameter regulation mode, the adjustment efficiency is greatly improved. The notch width consistency is improved by 65%, the fluctuation range is controlled within + / -0.02 mm, the end face roughness is reduced by 52%, the Ra value can reach 3.2 [mu] m or below, and the problems of thick plate slag residue, thin plate overburning deformation and the like are effectively solved.

The invention provides a wide graded gravelly soil catastrophe early warning method and system, and the method comprises the steps: obtaining the multi-source state data of a soil body, carrying out the data dimension reduction and analysis of the multi-source state data, and screening out a key disaster-causing factor set representing the wide graded gravelly soil catastrophe process; simulating a soil catastrophe evolution process on a mesoscopic numerical model by adopting a coupling simulation technology based on the key disaster-inducing factor set, and extracting a quantitative early warning index corresponding to the key disaster-inducing factor set from the evolution process; performing data-level fusion on the real-time monitoring values of the factors in the key disaster-inducing factor set obtained by the plurality of monitoring points by adopting a self-adaptive weighted average algorithm to generate a fused data sequence; based on the quantitative early warning index and the fusion data sequence, fusion calculation is carried out through fuzzy comprehensive evaluation, and a comprehensive risk assessment result is obtained; and determining a risk level according to the comprehensive risk assessment result, thereby effectively avoiding loss caused by lagging.

The invention belongs to the technical field of hydroelectric generating set fault intelligent diagnosis, and particularly discloses a hydroelectric generating set fault diagnosis method and device, electronic equipment and a storage medium. The method comprises the following steps: performing data preprocessing on multi-source vibration signal data of a hydroelectric generating set to be detected to obtain preprocessed multi-source vibration signal data; performing data dimension reduction on the preprocessed multi-source vibration signal data to obtain vibration signal feature data of a target dimension; inputting the vibration signal feature data of the target dimension into a unit fault diagnosis model to obtain a fault type of the to-be-detected hydroelectric generating set output by the unit fault diagnosis model; the unit fault diagnosis model is obtained by training according to a vibration signal feature data sample corresponding to the multi-source vibration signal data sample and a corresponding fault type label; the data of the target dimension and the target model parameters in the unit fault diagnosis model are determined through a parameter optimization method. According to the invention, the fault diagnosis precision of the hydroelectric generating set can be greatly improved.

The invention provides a data tracing method, device and equipment, a medium and a program product, which are applied to the technical field of data processing, and the method comprises the following steps: obtaining an original database and to-be-traced data; the method comprises the following steps: carrying out data dimension reduction on data to be traced to obtain first dimension-reduced data, carrying out feature extraction on the first dimension-reduced data to obtain a first data feature matrix, carrying out data dimension reduction on N first data in an original database to obtain N second dimension-reduced data, carrying out feature extraction on the second dimension-reduced data to obtain a second data feature matrix; n second data feature matrixes are obtained; based on the first data feature matrix and the N second data feature matrixes, second data similar to the to-be-traced data is determined from the N first data, a traceability database is constructed, and the traceability database comprises the second data; and matching the data to be traced with the data in the traceability database. Through the method provided by the invention, the accuracy of data tracing can be improved.

This invention relates to the field of wind power generation technology and discloses a method, device, equipment, and medium for fault assessment of wind turbine transmission systems. The method utilizes a data dimensionality reduction algorithm to reduce the dimensionality of multidimensional raw data, retaining core distinguishing features and simplifying calculations. Then, a data clustering algorithm is used to intelligently classify operating states, and a preliminary fault mode mapping is constructed by combining historical faults. Subsequently, the rationality of clustering is verified using signal source correlation coefficients, and long-term historical operating data is filtered through historical data matching rates to reduce the risk of misjudgment. Next, based on time series analysis and degradation path analysis, the coupling relationship between vibration trend slope, temperature accumulation offset, and torque decay cycle is obtained. Finally, core features are extracted through convolutional neural networks to accurately output the probability distribution and specific location of fault occurrence, thereby improving the accuracy of fault identification in the transmission system of offshore wind turbines and the ability to predict component performance degradation.

The invention relates to the technical field of underwater acoustic signal processing, in particular to a robust adaptive beam forming method based on prior information. Comprising the steps of performing dimension reduction processing on a data covariance matrix according to dimension reduction quantity parameters in historical data to obtain a data dimension reduction covariance matrix, and solving a data dimension reduction beam forming expression; deriving a preset value range of the loading amount according to the data dimension reduction covariance matrix; setting an initial value of a loading amount according to an optimal loading amount parameter in the historical data of the wave beam, and solving a current optimal loading amount; and if the current optimal loading amount is within the preset value range, solving a driving vector formed by the robust adaptive beam. According to the method, the data dimension reduction order and the initial value in the iteration process are set by utilizing the historical optimal data dimension reduction amount and the historical optimal loading amount, so that the convergence speed can be accelerated and the calculation complexity can be reduced. Meanwhile, the value range of the loading capacity is deduced to judge whether the diagonal loading capacity is effective or not, and divergence of the iteration process is avoided.

The optical push-scan positioning method for the reactor insulation fault comprises the following steps: supporting a beam to periodically push-scan under the actuation of a transmission device; collecting the light intensity distribution of a circular area on the bottom plane of the reactor under different light sources, extracting the gray image of the circular area, and obtaining the data matrix of the circular area; based on image transformation, the data matrix of the circular area is unfolded along the radial direction, and is converted into a rectangular data matrix through data filling; after unfolding, the circumference of the circle becomes the length of the rectangle, the radius of the circle becomes the width of the rectangle, the data of the circle center is changed into the length of the other side of the rectangle through data replication, and the data matrix of the circular area becomes the rectangular data matrix, and each column of the rectangular data matrix is the data along the radius of the data matrix of the circular area; based on the principal component analysis method, the data of the rectangular data matrix is analyzed to extract feature points, and multiple SiC ultraviolet solar blind sensor probes are linearly and non-uniformly arranged on the supporting beam according to the positions of the feature points.

The invention discloses a digital factory intelligent control method based on AI, and belongs to the technical field of intelligent control. The method specifically comprises the following steps: S1, multi-source data acquisition: acquiring equipment operation, production process, quality detection, environment and energy consumption data; s2, multi-source data preprocessing: performing cleaning and normalization preprocessing on the acquired multi-source data to generate an original feature data set; the three-level feature extraction mechanism breaks data islands and mines high-order semantic information, the first-level features achieve original data dimension reduction and information concentration, the second-level features capture cross-dimension causal association, the third-level features are converted into decision-level knowledge, four types of models of production scheduling, quality prediction, fault early warning and energy consumption optimization are integrated, the third-level high-order features are used as unified input, and the three-level high-order features are used as unified input. And for example, when the fault early warning model suggests'equipment load reduction ', the production scheduling model can synchronously adjust other equipment loads to guarantee the construction period.

The invention provides a height-configurable radar two-dimensional CFAR detection circuit system. The height-configurable radar two-dimensional CFAR detection circuit system comprises a target detection processor and a two-dimensional CFAR target detection hardware accelerator, the target detection processor is used for sending distance-Doppler matrix data, performing parameter configuration on the accelerator through a configuration interface, and receiving a detection result of the accelerator; and the two-dimensional CFAR target detection hardware accelerator comprises a frame data dimension reduction module and a sliding window target detection module, and is used for receiving the distance-Doppler matrix data, carrying out frame data dimension reduction processing and sliding window target detection according to the configuration parameters, and outputting a target detection result. The beneficial effects of the invention are that through cooperation of the target detection processor and the two-dimensional CFAR target detection hardware accelerator with a specific dual-module architecture, low power consumption and high real-time performance are ensured, and highly configurable adaptation of multi-scene radar parameters is realized.

The invention relates to the field of reliability analysis of aircraft hydraulic systems, in particular to a two-stage residual life prediction method of an aircraft hydraulic system based on data dimension reduction, which comprises the following steps: acquiring state data of the aircraft hydraulic system and preprocessing the state data; sequentially carrying out principal component analysis and kernel principal component analysis on the preprocessed data to obtain dimension reduction data; establishing a dual-target prediction model; inputting the trend term and the periodic term into a double-target prediction model to obtain a preliminary life prediction value; inputting the dimension reduction data into a residual prediction network to obtain a life residual prediction value; the residual prediction network is composed of a one-dimensional convolutional neural network and a bidirectional long-short term memory network which are connected in sequence; adding the initial life prediction value and the life residual prediction value to obtain a final residual life prediction value; according to the method, the residual life prediction efficiency and robustness of the aircraft hydraulic system can be improved.

According to the actuator fault diagnosis method and device based on the Spark MLlib and the medium, the typical fault mode of the electromagnetic valve is systematically analyzed on the basis of a Spark MLlib machine learning framework. Multi-source sensing data in a working state of an electromagnetic valve is collected, data preprocessing is firstly carried out to eliminate noise interference and dimensional difference, then fault feature vectors are extracted by adopting a wavelet packet decomposition algorithm, an energy feature set capable of effectively representing different fault types is constructed, and key fault information is reserved while data dimension reduction is realized. According to the method, a random forest multi-classification model is realized based on Spark MLlib, hyper-parameter optimization is performed through grid search, the performance of the model is evaluated by adopting a K-fold cross validation method, and the finally obtained fault classification accuracy reaches 96.2%.

The application relates to the technical field of model training, and discloses a lung cancer screening model training method and device based on gene transcriptome data, equipment and a medium. A basic model and reference transcriptome data are acquired. Core feature recognition is performed on the reference transcriptome data to perform data dimension reduction on the reference transcriptome data, and feature gene data is obtained. The basic model is pre-trained by using the feature gene data, and a preliminary training model is obtained. The preliminary training model is fine-tuned by using the reference transcriptome data, and an intermediate training model is obtained. In the case that the intermediate training model meets performance verification conditions, the intermediate training model is output as a lung cancer screening model. The beneficial effect is that core feature recognition and data dimension reduction are performed on existing reference transcriptome data, feature gene data suitable for large model training is obtained and used for training, the advance and objectivity of the gene transcriptome data are effectively utilized, and the screening timeliness and accuracy of the lung cancer screening model are improved.

The embodiment of the invention discloses a single cell transcriptome data processing method and device, a parameter library and electronic equipment, and the method comprises the steps: obtaining a common parameter, the common parameter comprises a reference feature gene set and a reference association relationship between an original feature and an extracted feature, the reference feature gene set comprises a plurality of feature genes, and the reference association relationship comprises a reference association relationship between the original feature and the extracted feature; the reference association relationship is used for dimension reduction processing of a gene expression condition; based on the reference feature gene set and the single cell transcriptome data of the to-be-detected sample, determining the gene expression condition of the feature gene in the to-be-detected sample; on the basis of the gene expression condition of the feature gene in the to-be-detected sample and the reference association relationship, performing first data dimension reduction processing to obtain a first dimension reduction result; wherein the to-be-detected sample and the common parameters belong to the same biological tissue type. By adopting the embodiment of the invention, the computing resource demand can be effectively reduced, and the data processing efficiency is improved.

The application relates to an abnormality detection method based on a hyperspectral image. The main body is based on a space-spectrum combined feature extraction method of filtering and low-rank decomposition to perform abnormality detection on the hyperspectral image. The specific method comprises the following steps: firstly, in the spatial dimension, a reduced dimension image is obtained through a data dimension reduction and eigenvalue weighted fusion method, and then an improved spatial filtering method is used to extract the spatial features of the image to obtain an initial spatial feature image. In the spectral dimension, a background reconstruction image of the approximate background is obtained by using a Tucker decomposition method on the original hyperspectral image, and a background dictionary of the image is obtained by using an improved k-means clustering method, then the background dictionary is input into a low-rank decomposition model to obtain a sparse matrix, and an initial spectral feature image is obtained, finally, the initial spectral feature image is fused with the spatial feature image to realize abnormality detection.

The invention provides a method for data reduction in the context of spectral CT imaging. The method comprises reducing the number of spectral channels of the acquired CT data by a reduced set of synthetic energy channels, each from a weighted combination of the original measured energy channel. The weights are selected in an optimization process in which an error metric associated with an output of a material decomposition procedure to be applied to the CT data is estimated, and the weights adjusted to minimize a value of the error metric. The error might for example be a noise estimate, and / or a bias estimate.

The invention relates to the technical field of data processing, in particular to a power grid load high-precision prediction method based on big data analysis and a storage medium. The method comprises the following steps: acquiring meteorological data and power grid load data; determining a period cooperation coefficient based on the period difference of the spectrogram peak values of the meteorological data and the power grid load data; screening a periodic point based on the periodic cooperation coefficient, and obtaining a periodic prediction value based on the meteorological data of the periodic point and the power grid load data; constructing a contribution function based on the power grid load data and the meteorological data, and determining the contribution degree based on the partial derivative and the maximum difference of the meteorological data; constructing an optimal function based on the contribution degree to perform dimension reduction on the meteorological data; determining a non-periodic predicted value based on the dimension-reduced meteorological data and the power grid load data; and obtaining a predicted value of the power grid load data based on the periodic predicted value and the non-periodic predicted value. The prediction precision of the power grid load is improved.

The invention discloses an auxiliary radio frequency circuit modeling method based on machine learning, and belongs to the technical field of radio frequency circuits. Comprising the following steps: acquiring design parameters of a radio frequency circuit by utilizing a Latin hypercube sampling method to obtain sample data, performing electromagnetic simulation on the sample data through electromagnetic software to obtain corresponding electromagnetic response, obtaining a high-simulation-precision electromagnetic response data set by matching the Latin hypercube sampling method with the electromagnetic software, and taking the sample data as input characteristics to obtain a high-simulation-precision electromagnetic response data set; carrying out dimension reduction processing on sample data by utilizing a Pearson correlation coefficient in cooperation with a kernel PCA method, building a neural network model, optimizing a hidden layer in the neural network model and the number of neurons of the hidden layer by utilizing a whale optimization algorithm to obtain a perfect neural network model, and taking a global optimal solution obtained by a WOA algorithm as an initial parameter of the neural network model; and the initial parameters are processed to obtain initial parameters, and the neural network model is trained through the initial parameters. According to the invention, the neural network model with high precision is built in a short time.

This invention discloses a method for acquiring remote fish spoilage prediction data based on edge computing. The steps are as follows: acquiring fish spoilage gas detection data for training a variational autoencoder (VAC) model; adding spoilage labels to the fish spoilage gas detection data for training the encoder neural network structure of the VAC as a classifier model; transferring the classifier model to a new VAC and then retraining the decoder; running the VAC on the edge computing device of the fish spoilage detection system, simultaneously generating a spoilage prediction result and dimensionality-reduced data; sending the generated data to a remote server via the BeiDou messaging system; and deploying a decoder on the remote server to acquire the original detection data. The method provided by this invention can simultaneously achieve data dimensionality reduction and spoilage prediction functions at the fishing vessel and edge computing device, enabling remote monitoring; furthermore, the stored data can serve as a reserve for higher-precision prediction.

This invention provides an intelligent evaluation method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks, belonging to the field of rockburst hazard evaluation technology. An evaluation index system is established and data is collected by combining rockburst hazard influencing factors, drill cuttings monitoring, critical stress index monitoring, and actual field conditions. Principal component analysis (PCA) is used to simplify the original evaluation index data, resulting in comprehensive evaluation index data containing information from the original evaluation indicators. The comprehensive evaluation index data is divided into a training set and a test set. A probabilistic neural network (PNN) is used to train the evaluation model on the data in the training set, and then the performance of the evaluation model is tested using data from the test set, and its accuracy is calculated. If the accuracy is greater than or equal to 90%, the evaluation model is considered acceptable; if the accuracy is less than 90%, it is considered unacceptable, and the PNN's smoothing factor needs to be modified and retrained until the accuracy of the evaluation result meets the set value.

The invention discloses a clustering method. The method comprises the following steps: respectively carrying out one-dimensional data clustering on dimensions contained in data; and based on the relationship between the cluster of the data and the one-dimensional data cluster, obtaining the cluster of the data according to the one-dimensional data cluster. According to the method and the device, the problem of dimension disaster in related technologies is solved, so that the clustering performance during high-dimensional data clustering is greatly improved, data dimension reduction is avoided, a series of problems caused by dimension reduction are solved, and the clustering effect is improved.