56 results about "Negatively associated" patented technology

The invention provides an industrial product quality inspection model building method, a medium and a system, and belongs to the technical field of computer models.The industrial product quality inspection model building method comprises the steps that firstly, multi-source data including process parameters, visual features, thermal imaging features and the like are obtained, and the multi-source data are subjected to data cleaning, grouping sorting and feature normalization processing; and constructing a quality data vector. And then calculating a batch quality data curve and an intra-batch interaction index based on the vector, and modeling through a negative correlation coupling neural network. The network comprises two core coupling equation sets which are responsible for feature extraction fusion and mass transfer effect analysis respectively, and effective decoupling and fusion of information are achieved through a shared feature layer and negative correlation constraints. And finally, through training and verification, a quality inspection model capable of accurately quantifying the relevance of the quality characteristics of the products in the batches is constructed, and the technical problem that the relevance and the transfer effect of the quality characteristics of the adjacent products in the industrial product batches are difficult to accurately quantify and predict is solved.

The invention relates to the technical field of data processing, in particular to an anomaly detection method and system for a refrigeration valve, and the method comprises the steps: obtaining the preprocessed operation data of the refrigeration valve, and for any data point in the operation data, calculating the sparse degree of the data point according to the distribution state of the data point in a local region; meanwhile, constructing a time sequence adjacent reference set of the data points, and obtaining an initial abnormal score of each data point in the adjacent reference set by using an LOF algorithm; constructing a trend anomaly degree according to the initial anomaly score of the data point in the time sequence adjacent reference set and the change of the data value; a preset initial abnormal threshold value is adjusted based on the trend abnormal degree and the sparse degree to obtain a dynamic threshold value, the dynamic threshold value is in negative correlation with the trend abnormal degree, and the operation data of the refrigeration valve are judged based on the dynamic threshold value. The method has the effect of improving the abnormality detection accuracy of the refrigeration valve.

The invention relates to the field of data processing, in particular to a data management method and system for Internet of Things equipment based on big data. The method comprises the following steps: acquiring an Internet of Things equipment data sequence; obtaining the occurrence probability of each value in the Internet of Things equipment data sequence, recording the occurrence probability as a comprehensive probability, and obtaining the distribution preposition degree of any value; calculating an adjustment coefficient of each value, wherein the adjustment coefficient is in positive correlation with the distribution preposition degree; setting a segmentation proportion of each value according to the adjustment coefficient and the comprehensive probability, wherein the segmentation proportion is negatively correlated with the adjustment coefficient and negatively correlated with the comprehensive probability; and coding control is carried out according to the segmentation proportion to realize data storage management. The segmentation proportion is adjusted according to the data arrangement information, and the compression effect is improved.

To diversify environments for evaluating measurement waveforms of biological parameters acquired from an object person.SOLUTION: An inference device 11 acquires first data D1 corresponding to a measurement waveform of biological parameters of an object person 20, infers the probability of the value of the biological parameters being measured correctly, and generates second data D2 in which the probability is added to the first data D1. A processing device 12 calculates a value score of the measurement waveform by applying a predetermined evaluation function to the second data D2 in which the probability exceeds a threshold, and generates third data D3 in which the value score is added to the second data D2. The value score has a positive correlation for the probability, and a negative correlation for the time elapsed from the measurement of the biological parameters. The processing device 12 acquires an index value showing a risk of a sudden change in the condition of the object person 20 on the basis of at least one piece of the third data D3 in which a higher value score is added. The processing device 12 changes at least one of the threshold and the evaluation function on the basis of the index value.SELECTED DRAWING: Figure 1

A data analysis computer system is provided that receives a timeseries dataset and generates implied data from the dataset. The dataset is further vectorized to reduce the dimensionality of the data. Users provide input to identify windows of data that either positively or negatively correlate to instances of a given type of occurrence within the data. The user defined windows are converted to fixed sized windows and a machine learning algorithm constructs a model from the data. The model is used to predict instances of the given type of occurrence in newly received data. Validation of the predications may be performed.

The application discloses a kind of based on radar chart's hole making quality evaluation method, system and storage medium, belong to the technical field of hole making quality evaluation, determine N features data of influencing hole making quality, and corresponding N-dimensional radar chart is drawn as radar chart evaluation model;Obtain the feature data of hole making, adopt box plot method to correct.Based on the feature data after correction, draw closed radar chart;The area of the radar chart drawn is calculated, and the quality of hole making is measured based on the size of area;If N feature data is all negatively correlated with the quality of hole making, then the smaller the area of the radar chart drawn, the higher the quality of hole making;If N feature data is all positively correlated with the quality of hole making, then the larger the area of the radar chart drawn, the higher the quality of hole making.The application realizes multidimensional feature correlation evaluation, avoids the difference caused by subjective evaluation, intuitively compares the quality of hole making by radar chart area, quickly locates the advantages and disadvantages of hole making quality, with good practicability.

The invention provides a product recommendation method and device, equipment, a medium and a product, and can be applied to the field of artificial intelligence. The product recommendation method comprises the steps of obtaining product resource data of preferred products of a target object and candidate resource data of candidate products, wherein candidate text data in the candidate resource data is used for describing product operation activities of the candidate products in a preset historical period; processing the product resource data and the candidate resource data based on a correlation algorithm to obtain a correlation degree; determining an intermediate product which does not meet a preset correlation degree threshold value from the candidate products based on the correlation degree; performing semantic understanding on the candidate text data by using a large language model to obtain a correlation influence relationship between the preference product and the intermediate product; and under the condition that the correlation influence relationship meets a preset negative correlation condition, determining a target product from the intermediate products, and pushing the target product and the preference product to the target object.

The embodiment of the invention provides a method for training an insurance amount prediction model. The method comprises the following steps: acquiring a labeled sample set containing insurance element information; and performing first training on a generator of the prediction model, and a first regression device and a second regression device which are initialized respectively based on the annotation sample. The generator is used for generating a first feature according to an input sample; the first regression device and the second regression device are used for outputting predicted amounts according to the whitening features of the first features. And inputting a first sample in the unlabeled sample set into the prediction model to obtain first / second prediction results output by the first / second regression device. The first / second regression device is subjected to first updating with the purpose of minimizing the difference loss, and the difference loss is negatively correlated to the relative difference between the first / second regression device; the relative difference at least comprises the difference between the first prediction result and the second prediction result. And performing second updating on the generator by taking the relative difference minimization as a target.

Provided is an information processing apparatus that performs processing for presenting a relationship between variables in multivariate analysis. The information processing apparatus includes a detection unit that detects a combination of two variables having a characteristic relationship in multivariate analysis, and a presentation unit that presents information regarding the characteristic relationship between the two variables. The detection unit detects whether or not there is a characteristic relationship including at least one of a positive correlation, a negative correlation, or a non-linear relationship as the entire variables on the basis of the relationship between the explanatory variable and the explained variable for each of the two consecutive categories of the explanatory variable, and further quantifies the relationship between the explanatory variable and the explained variable as the entire variables.

The invention relates to the technical field of cooperative control, in particular to a cooperative optimization control method and system for the expansion degree and strength of flow-state solidified soil, and the method comprises the steps: designing a sample according to all influence factors, measuring performance data, and forming test data; constructing a correlation matrix through preprocessing such as abnormal value elimination, variable standardization and virtual variable conversion; screening significant influence items through a variance analysis model and quantifying weights; constructing a dual-output prediction model based on a random forest algorithm, and performing training verification optimization; taking the prediction model as a constraint, and combining economy and convenience to screen an optimal parameter combination; on-site trial mixing verification is carried out, deviation reaches the standard, parameters are determined, and otherwise, data are complemented to update the model to resolve. According to the method, the single factor and the interaction effect are effectively removed, the problem of collaborative prediction under negative correlation of the expansion degree and the strength is solved, collaborative optimization of the expansion degree and the strength is realized, and the scientificity and efficiency of construction parameter determination are improved.

The invention discloses an abnormal data detection method applied to a battery system and a related device, and relates to the field of clustering algorithms, and the method comprises the steps: obtaining a plurality of attribute parameter sets of the battery system, and determining a plurality of original clustering centers from the attribute parameter sets; after the clusters corresponding to the multiple original clustering centers are obtained, the distance between each attribute parameter set and the corresponding original clustering center is obtained for each cluster, the weight is determined according to the distance, and due to the fact that the weight and the distance are in negative correlation, the contribution degree of the attribute parameter set with the large distance in subsequent clustering center updating is automatically attenuated; re-determining an updated clustering center based on a weighting mechanism, and continuously optimizing the updated clustering center corresponding to each cluster through repeated iteration; in the iteration process, if the standard function value is smaller than a preset threshold value, it is indicated that the interior of each cluster meets the iteration condition; at the moment, based on the optimized clusters, the attribute parameter set which cannot be absorbed by any cluster is determined as abnormal data located outside the multiple clusters, and therefore abnormal data identification is achieved.

An image reconstruction method, comprising: reading raw data; reducing dimensionality of the raw data to obtain a one-dimensional vector; taking auto-correlation of the one-dimensional vector to obtain an auto-correlation matrix; performing eigen-decomposition of the auto-correlation matrix to obtain eigenvalues and eigenvectors of the auto-correlation matrix; sorting the eigenvectors according to their corresponding eigenvalues; taking auto-correlation of a subspace constituted by eigenvectors corresponding to the first N smaller eigenvalues to obtain a target subspace; defining grid positions of a spectral search in a phase encoding direction; for each phase encoding direction, performing the following spectral search steps: defining an array manifold corresponding to the phase encoding direction; obtaining a correlation coefficient of the array manifold corresponding to the phase encoding direction and the target subspace or obtaining a correlation coefficient of an inverse matrix of the array manifold and the target subspace, and then performing a negative correlation transformation on the correlation coefficient to obtain a spectrum; extracting elements from the spectrum to obtain a row of a reconstructed image; and outputting the reconstructed image.

The invention relates to a classification model training method and device, a classification method and device, electronic equipment and a program product. The method comprises the steps that training data are acquired, the training data comprise a plurality of training samples, and each training sample comprises a plurality of features and corresponding classification labels; the training data are input into a classification model based on preset target optimization logic for training, a trained classification model is obtained, the target optimization logic comprises a prediction error item, the prediction error item is determined based on a sample weight, and the sample weight is in negative correlation with the possibility that the sample is correctly classified. According to the scheme provided by the invention, the sample weight is negatively correlated with the possibility that the samples are correctly classified, so that the training process can give a higher focusing degree to the samples difficult to classify, and the classification effect of the classification model is improved.

The invention relates to the technical field of transformer fault prediction, in particular to an electric power engineering safety quality intelligent diagnosis system based on AI, and the system comprises an obtaining module which is used for obtaining a plurality of normal data, a plurality of fault data and a plurality of expansion data outputted through a generative adversarial network; the analysis module is used for acquiring an effective value and an ineffective value of each piece of expansion data; the determination module is used for determining a retention value of each piece of expansion data according to the effective value and the invalid value of each piece of expansion data; and the training module is used for constructing a training set based on the plurality of normal data, the plurality of fault data and the plurality of removed expanded data, and training according to the training set to obtain a fault prediction model, and the retention value and the probability that the corresponding expanded data is removed are in a negative correlation relationship. According to the method, redundant data and heterogeneous data in the extended data can be screened out, so that the fault prediction capability of the fault prediction model obtained by training is improved.

The invention belongs to the technical field of data processing, and discloses a multivariable time sequence anomaly detection system and method based on positive and negative correlation graph deviation, and the system comprises a variable embedding module which is used for embedding each variable; the positive and negative variable correlation graph construction module is used for acquiring correlation of all time sequence variables so as to construct a positive and negative variable correlation graph of the multivariable time sequence; the information propagation and prediction module is used for obtaining information aggregation representation according to a correlation graph between positive and negative variables, and further obtaining a predicted value of each variable; the anomaly detector is used for calculating an error between a predicted value and an actual value of each variable to obtain anomaly-related time and variables. According to the invention, by introducing the positive and negative correlation diagram deviation, the dynamic dependency relationship between variables can be more accurately represented, the anti-interference capability of the system is improved, the time sequence anomaly detection performance is further improved, and the anomaly detection requirement in a complex scene is effectively met.

A data analysis computer system is provided that receives a timeseries dataset and generates implied data from the dataset. The dataset is further vectorized to reduce the dimensionality of the data. Users provide input to identify windows of data that either positively or negatively correlate to instances of a given type of occurrence within the data. The user defined windows are converted to fixed sized windows and a machine learning algorithm constructs a model from the data. The model is used to predict instances of the given type of occurrence in newly received data. Validation of the predications may be performed.

The invention provides a product pushing method and device based on financial personality and electronic equipment, and the method comprises the steps: collecting a plurality of pieces of user feature data of a user under the condition of obtaining user authorization; a factor analysis method is adopted to determine scores corresponding to the user feature data in multiple dimensions, and multiple user feature scores are obtained; constructing a training data set by adopting a positive and negative correlation factor method based on the user feature score, and training a personality classification model according to the training data set; obtaining personality classification data, and processing the personality classification data by adopting the personality classification model to obtain a target financial personality; and determining a corresponding product pushing strategy according to the target financial personality, and performing product pushing according to the product pushing strategy. The method solves the problems that in the prior art, user positioning is carried out through a user portrait, analysis on user behavior psychology is lacked, positioning is inaccurate, and product pushing does not conform to user preferences.

The present invention provides an SRAM PUF key extraction method based on correlation judgment, comprising: an initial registration phase: powering on the SRAM, configuring and initializing variables; calculating the exclusive OR of a data block and its cyclic shifts of different points to obtain the number of cyclic shift points and the number of negatively correlated bits when the maximum negative correlation occurs; selecting N valid data blocks whose maximum negatively correlated bit numbers are greater than a set threshold; using N bits of hidden random numbers to control whether the valid data blocks are cyclically shifted according to the number of their maximum negatively correlated cyclic shift points; linearly encoding a randomly generated PUF output random number; performing an exclusive OR operation on the encoding result and the hidden random number to obtain auxiliary data; a reconstruction phase: powering on the SRAM, initializing variables, and obtaining a valid data block based on the auxiliary data selection; restoring an approximate random number of the hidden random number using the valid data block and the auxiliary data; performing an exclusive OR operation on the approximate random number and the auxiliary data to obtain a codeword before error correction; and performing linear encoding on the codeword to correct errors and restore the PUF output random number.

This application relates to the field of data processing technology, and in particular to a method and system for anomaly detection of refrigeration valves. The method includes: acquiring preprocessed operating data of the refrigeration valve; for any data point in the operating data, calculating the sparsity of the data point based on its distribution in the local area; simultaneously, constructing a temporal neighbor reference set for the data points, and using the LOF algorithm to obtain the initial anomaly score for each data point in the neighbor reference set; constructing a trend anomaly degree based on the initial anomaly scores of the data points in the temporal neighbor reference set and the changes in data values; adjusting a preset initial anomaly threshold based on the trend anomaly degree and the sparsity to obtain a dynamic threshold, wherein the dynamic threshold is negatively correlated with the trend anomaly degree; and judging the operating data of the refrigeration valve based on the dynamic threshold. This application has the effect of improving the accuracy of anomaly detection in refrigeration valves.

This invention belongs to the field of pipeline leakage detection technology, and relates to a method for identifying leaks in water supply networks used in urban lifeline projects. The method involves collecting pressure fluctuation data and synchronous flow data; calculating the difference between the pressure fluctuation data and a benchmark fluctuation template point by point to obtain a fluctuation residual sequence; extracting features from the fluctuation residual sequence to obtain spatial distribution characteristics; combining the spatial distribution characteristics and synchronous flow data to calculate the negative correlation between pressure drop and flow increase, as well as the time difference between pressure drop and flow increase; and marking suspected leakage features based on the calculation results. Finally, the method performs correlation analysis on the suspected leakage features based on pipe aging parameters and historical maintenance records to identify leaks. This invention uses a benchmark fluctuation template to strip away normal fluctuations to highlight weak abnormal signals, and integrates pipe aging parameters and historical maintenance records for multi-factor correlation confirmation. It progressively separates minute leakage features from complex background noise, effectively improving the accuracy of leak identification and reducing the false alarm rate.

In this image, data is input to an AI model by adjusting the length of the bars extending horizontally. For example, if the data input to the AI ​​model is the Iris dataset, the length and width of the sepals and petals of iris flowers are parameters, and the magnitude of these values ​​can be changed by manipulating the bars. A histogram corresponds to a combination of bar lengths and visually represents the correlation between each parameter. For example, if there are many low-brightness areas in the histogram, there is a positive correlation, and if there are few low-brightness areas, there is a negative correlation.

Embodiments of the present application disclose a kind of energy consumption prediction method, device, electronic equipment and storage medium of electrical object, comprising: according to the preset period, the multiple historical energy consumptions of electrical object are collected;Set the clustering target function with weight to the historical energy consumption clustering is carried out to be granulated and obtain granulated data, in the clustering target function, the distance of historical energy consumption to cluster center is negatively correlated with the weight;The granulated data is degranulated, and the data after degranulation is obtained;Using the data after degranulation establishes fuzzy rule model to predict the energy consumption of the electrical object.The embodiments of the present application, since in the clustering target function, the distance of historical energy consumption to cluster center is negatively correlated with the weight, can be dynamically adjusted in the clustering process by weight to clustering target function, further refine the distance of different historical energy consumptions to each cluster center, improve the clustering effect of historical energy consumption, ultimately improve the accuracy of energy consumption prediction of electrical object.

The present application relates to the technical field of retrieval, more particularly, the present application relates to a kind of property information cross-modal retrieval method and system based on knowledge graph, comprising: the multi-modal attribute graph of static node and dynamic event node is constructed, each node in the graph is given source label according to data source, and text information and image information in the graph are processed into feature vector;The source entropy confidence of each node is calculated, and the source entropy confidence and the historical data dispersion of the corresponding source of the node in the preset time window are negatively correlated.The present application realizes the quantitative evaluation of the objectivity of different sources (such as IoT and artificial record) by constructing multi-modal attribute graph and calculating source entropy confidence, comprehensively considers historical dispersion, text and picture self-consistency and credibility factor, effectively suppresses subjective noise according to low-entropy high-power principle.

The present application relates to the technical field of fault early warning, more specifically, the present application relates to a kind of cutting fluid filling equipment's fault automatic early warning method and system, comprising: the acquisition cutting fluid filling loop's multiple-source flow variable data, include pressure, flow and fluid density data;Based on the pressure data and flow data in the multiple-source flow variable data.The present application constructs impedance factor and the discrete significance of statistics by combining the principle of fluid mechanics, quantifies the severity of failure, and establishes a negative correlation mapping between fault significance and algorithm calculation dimension, improves the extended isolation forest algorithm, automatically reduces the feature dimension to reduce the calculation redundancy when facing serious failure, maintains high-precision in weak failure, thereby solving the problem of slow response and wasted computing power of existing fixed dimension algorithm in high-speed filling scene, realizing real-time, low-latency and accurate early warning.

The invention provides a data table storage method and device and a data storage system.The method comprises the steps that related information of a data table is obtained, and the related information comprises one or more of the number of calling times, the urgency degree of service requirements, the proposing time of the service requirements and the data size; the comprehensive score is the score of the importance of the data table, the calling frequency and the comprehensive score are in positive correlation, the emergency degree of the service demand and the comprehensive score are in positive correlation, and the distance between the proposing time of the service demand and the current time and the comprehensive score are in negative correlation; the data volume and the comprehensive score are in negative correlation; sorting all the data tables according to the comprehensive scores to obtain sorted data tables; and storing all the sorted data tables into a database according to the sequence of the sorted data tables. According to the method, the problem that in the prior art, when data is stored in a database, due to the fact that the data size is large, the efficiency is poor in the whole database storage process is solved.

The present disclosure relates to a training method, device, and computer-readable storage medium for a classification model, and relates to the field of artificial intelligence technology. The method includes: dividing samples for training into multiple sub-batches, wherein each sub-batch includes one or more samples; inputting the samples of each sub-batch into a classification model, determining a gradient vector of the classification error corresponding to each sub-batch relative to an initialization parameter of the classification model; determining a value of an objective function based on the gradient vector corresponding to each sub-batch, wherein the objective function is negatively correlated with the classification loss of the classification model; if the value of the objective function does not reach a maximum value, adjusting the initialization parameters of the classification model according to the objective function; until the value of the objective function reaches a maximum value, using the initialization parameters of the classification model as optimized initialization parameters, and training the classification model with the optimized initialization parameters.

The invention relates to the technical field of piston life prediction, and particularly discloses a data-driven piston life prediction method, which comprises the following steps: establishing a piston life prediction model, and periodically inputting real-time characteristics into the piston life prediction model to obtain the residual life W of a piston; if the residual life W is smaller than a preset residual life threshold value, obtaining a numerical value of the characteristic in the verification period and recording the numerical value as a historical value; sorting the historical values of the single types according to a time axis sequence to obtain a first sequence, and calculating a verification rate; determining a positive characteristic and a negative characteristic according to a positive correlation or negative correlation relationship between the characteristic and the residual life of the piston; determining whether the positive feature and the negative feature are abnormal features based on the verification rate and the historical value; and judging whether to send piston life early warning or not according to the total number of the abnormal characteristics. According to the invention, the accuracy of sending the piston life early warning is improved.

The invention belongs to the technical field of data mining, and particularly provides a stock trend prediction method and system based on a space-time hypergraph quadruple attention network, and electronic equipment, and the method comprises the following steps: 1, carrying out the preprocessing and feature extraction of stock historical price data, and obtaining stock embedding; 2, the stocks are divided into positive correlation and negative correlation types, and information from the two types is aggregated through a weighted graph attention mechanism; and 3, introducing an industry hypergraph and a fund holding hypergraph, and processing the dominant relationship among the stocks through a three-layer attention network. And 4, extracting complementary information embedded in different stocks through two-stage attention, and fusing the complementary information. And step 5, carrying out three-classification after passing through a full connection network. According to the method, the deep data mining technology is adopted, the space-time hypergraph quadruple attention network is introduced, the stock is predicted by combining the market time dynamic state and the data relation of the multi-dimensional stock, the prediction capability is enhanced, and robustness and accuracy are achieved.

The embodiment of the invention provides a data set construction method, a model training method, an object recommendation method and a content recommendation method, and the data set construction method comprises the steps: obtaining a return visit behavior set of a target user, the return visit behavior recording a user identifier, an object identifier, a click tag for clicking an object or not, and an access timestamp; determining a return visit interval according to the access timestamp; respectively generating a positive correlation label and a negative correlation label according to the click label and the return visit interval; and based on the return visit behavior set, the positive correlation tag and the negative correlation tag, constructing a data set of a training object sorting model. Accurate interest preference modeling of the object sorting model in the cold start stage can be effectively supported, and the accuracy of object recommendation and the user activeness are improved.

A method for improving reduction of negative correlation noise (anti-correlation noise) artifacts in spectral CT imaging. A first set of embodiments employs operations performed in a pre-decomposed projection domain by first decomposing acquired projection data (optionally only with low resolution decomposition), a respective set of decomposed data values for each sinogram voxel is mapped to a known pre-calculated statistical distribution of anti-correlation noise associated with that set of decomposed data values, and then a smoothing operator is applied that is tuned to reduce all noise variance values below a set threshold. Another set of embodiments includes detecting photon detector counts (low count readings) below a particular threshold, and then determining a reliability index for each sinogram voxel with the set of detected low count readings. Smooth operators are utilized to operate on those readings in the sinogram that are below threshold reliability. This is based on a pre-calculated probability distribution of different possible sets of probe count readings that is based on a theoretical probability of each combination of the read counts of a hypothetical attenuation distribution for a given scan volume.