13 results about "Data averaging" patented technology

The invention relates to a city development talent demand prediction method and device, and a storage medium. The method comprises the following steps: selecting city research time nodes, input elements and output elements; collecting corresponding input element data and output element data of the city at the research time node; based on the input element data and the output element data, adopting a Cobb-Douglas production function to determine the output elasticity of each input element of the city; determining marginal output data and average output data of each input element according to the output elasticity; drawing a total output curve, an element marginal output curve and an average output curve of the city according to the total output data, the marginal output data and the average output data of each research time node of the city; according to the curve position relation, the urban talent development stage is determined, and a prediction result of the talent demand is generated. According to the method, the effects of capital, labor force and technical progress factors in the production process are described from the macroscopic perspective, and the influence of the talent number and quality on urban development is quantified.

The invention belongs to the field of GPU multi-card interconnection communication, and relates to a full-exchange GPU communication method and system. According to the method, the data stored in each GPU is averagely divided into N blocks, and each block of data corresponds to a specific data block needing to be sent to the target GPU; constructing a data symmetric matrix, and determining a total communication round according to the total number N of the GPUs; in each round of communication, two-way data transmission is carried out between the two GPUs based on a pairing rule of a symmetric matrix; and through multi-round iteration, complete exchange transmission of all data blocks is completed, so that each GPU obtains a complete data set from all GPUs. According to the method, a connection topology scheme of the GPUs does not need to be changed at all, the symmetric matrix of the GPUs only needs to be generated according to the number of the GPUs, and the condition that the GPUs communicate with other GPUs at the same time is reduced under the condition that the data transmission quantity is not increased, so that the congestion possibility of the GPUs is reduced, and the communication performance is improved.

The invention relates to the technical field of solder paste management, and provides a solder paste multi-dimensional data fusion management method and system, and the method comprises the steps: creating a quality label corresponding to each batch of solder paste; continuously receiving temperature and humidity data from the temperature and humidity sensors in all the areas in the refrigerated cabinet, and preprocessing the temperature and humidity data to correct the temperature and humidity data which are higher than the data average value of all the areas but do not reach a preset abnormal alarm threshold value; calculating and updating a quality label according to the preprocessed temperature and humidity data, solder paste material characteristics and storage environment parameters; evaluating the sensitivity of each production task to the solder paste quality in the production process to obtain a sensitivity level; according to the sensitivity level and the updated accumulative microstress damage value, the solder paste batch is dispatched; and according to the updated accumulative microstress damage value, issuing a process adjustment suggestion to production equipment. The method has the effect of improving the management precision and efficiency of the full life cycle of the solder paste.

The invention discloses a secret key using method and device and a storage medium. The method comprises the steps that n data processing processes are created and used for conducting concurrent processing on data; obtaining n unused key files with the length of L from the key files with the total length of M according to the data average length L within a past period of time T, and sequentially and correspondingly distributing the n key files to corresponding data processing processes; and distributing the to-be-processed data to the corresponding data processing process for encryption processing. According to the invention, a segment of key is preset in each data processing process, so that the data processing process can immediately execute encryption operation when receiving the to-be-processed data without waiting to request to distribute the key to a key file; particularly, under the condition that the length of the data to be processed is smaller than the length of the preset key, encryption can be completed by directly using the preset key, the encryption preparation time is remarkably shortened, and the real-time performance of data processing is improved.

The application provides a health diagnosis method suitable for a GNSS receiver, and the method comprises the following steps: determining to-be-diagnosed sites, and acquiring calculation data and original data of GNSS receivers of each to-be-diagnosed site within a preset time period; acquiring a standard deviation of the calculation data of the GNSS receivers of each to-be-diagnosed site within the preset time period; acquiring a data average value of each quality parameter in the original data of the GNSS receivers of each to-be-diagnosed site within the preset time period; acquiring a correlation coefficient of each quality parameter and the standard deviation; acquiring quality parameters corresponding to the correlation coefficients in the top 50%; obtaining a normalized data average value of each quality parameter; and judging an influencing factor of GNSS receiver accuracy based on the normalized data average value of each quality parameter, so as to complete the health diagnosis of the GNSS receiver. The application can solve the technical problem that it takes too long to judge the authenticity of ground surface deformation by using manual work in the prior art.

Aspects of the disclosure relate to using machine learning models to automatically deallocating overallocated database storage. A computing system may retrieve database state data that indicates current states of databases. Further, the computing system may retrieve average storage usage thresholds and peak storage usage threshold for the databases. The average storage threshold for the databases may indicate an average amount of storage used by each database. The peak storage usage threshold for the databases may indicate an amount of storage that is allocatable without causing an abnormal termination. Based on inputting the database state data, the average storage usage threshold, and the peak storage usage threshold into the machine learning models, an amount of overallocated storage of the databases may be determined. Further, the overallocated storage may be deallocated from the databases based on configurations of each of the databases.

The application provides an on-board abnormal data determination method and device and electronic equipment. The method comprises the following steps: selecting target on-board data corresponding to the number of independent variables of a data table; determining an on-board data average curvature sequence and an on-board data Gaussian curvature sequence corresponding to the target on-board data; determining an average curvature non-abnormal value interval and a Gaussian curvature non-abnormal value interval based on a relative deviation parameter, average curvature quartiles and average curvature quartile distance, Gaussian curvature quartiles and Gaussian curvature quartile distance; determining an average sequence abnormal data and a Gaussian sequence abnormal data based on an absolute deviation parameter, the average curvature non-abnormal value interval and the Gaussian curvature non-abnormal value interval; and combining the average sequence abnormal data and the Gaussian sequence abnormal data to obtain abnormal data of the target on-board data. The on-board abnormal data determination method, device and electronic equipment solve the problems of long time consumption, low reliability and low efficiency in checking on-board abnormal data by manual means.

The invention provides a covert communication-oriented communication and inductance integration performance optimization method. The method comprises the steps of configuring a covert communication-oriented low-altitude network model in an unmanned aerial vehicle relay scene; based on the low-altitude network model, determining performance indexes in the aspects of communication and perception; on the basis of the indexes, modeling is carried out on the covert communication-oriented communication and inductance integration performance optimization problem; and based on improved multi-operator differential evolution, solving the modeled optimization problem. According to the method, a multi-index weighted optimization problem of a joint security communication reachable rate, a transmission data average information age and a beam pattern error facing covert communication is established; the provided optimization problem solving algorithm based on improved multi-operator differential evolution has strong global search capability, can effectively avoid falling into a local optimal solution, also has an adaptive parameter adjustment mechanism, can dynamically adjust the parameters of the algorithm according to the evolution process, and improves the convergence rate and robustness of the algorithm.

The invention discloses an AI adaptive feature extraction method for real-time streaming data of an internet platform, and particularly relates to the technical field of data processing. According to the method, quality parameters such as feature effective rate and feature stability and cost parameters such as average extraction time consumption of single data, comprehensive resource occupancy rate and benefit ratio are integrated, and a policy effect estimation system is constructed through normalization processing and weighted summation, so that a policy effect estimation value capable of comprehensively reflecting policy technology quality and commercial value is obtained; scientific quantitative evaluation on the effect of the feature extraction strategy is realized, and the problem that whether the strategy needs to be optimized or not is difficult to accurately judge due to isolated attention to single dimension in the prior art is solved.

This invention discloses a bridge construction deflection sensitivity analysis method, comprising the following steps: Step S1, establishing a historical database; Step S2, standardizing each parameter in the historical database; Step S3, based on the standardized parameter data, taking the average of all data for that parameter, and generating random numbers with a sample size of N from the parameter distribution; Step S4, integrating a hybrid learner; Step S5, establishing a three-dimensional deflection model of the main beam using the hybrid learner; Step S6, obtaining multiple deflection values ​​as sample output values ​​using the arithmetic mean median method; Step S7, calculating the conditional expectation of each parameter with respect to the sample output values; Step S8, calculating the magnitude of the influence of the parameters on the main beam deflection using the Sobol sensitivity analysis system. This invention, by analyzing the influence of each parameter on the main beam deflection, guides the strict observation and control of parameters with a significant impact during construction, thereby avoiding resource waste and reduced construction efficiency.

The invention provides an intelligent electrocardiosignal classification method based on a mixed cross-correlation distance, which comprises the following steps of: 1, performing dimensionality reduction compression on an original electrocardiosignal time sequence with the length of L by adopting a segmented aggregation approximation PAA algorithm, and calculating a data average value in each segment to obtain a compressed sequence represented by PAA; 2, calculating a mixed cross-correlation distance between the to-be-measured data and the candidate data, wherein the distance is obtained by weighting and combining a global cross-correlation distance and a local cross-correlation distance through a harmonic parameter alpha; and 3, sorting the mixed cross-correlation distances in an ascending order, selecting the first K neighbors, and determining the final category of the electrocardiosignals to be detected according to a majority voting principle. According to the technical scheme, the calculation complexity is remarkably reduced while high classification precision is guaranteed, so that the requirements of a modern medical monitoring system for high precision, low delay and high robustness are met.

The invention discloses a geometric modeling method and system for a stator and rotor deformation contour of a large hydro-generator, and belongs to the field of hydro-generator vibration faults. The method comprises the following steps: acquiring original air gap data of each magnetic pole acquired by an air gap monitoring system at a plurality of angle monitoring points of an inner circle of a stator core, extracting the air gap data of the same magnetic pole at each monitoring point and averaging to obtain original data of an outer circle contour of a rotor magnetic pole; obtaining a rotor deformation contour according to the initial position of the deviation offset magnetic pole; an average value of air gap data of each monitoring point is obtained to obtain original data of an inner circle contour of the stator core, and after the original data is added with a rotor design standard outer circle contour size, a real deformation contour of the stator core is obtained through data expansion, fitting, tooth groove extension and Boolean operation. According to the method, the deformation characteristics of the stator and the rotor are precisely reduced, the process is standardized, different units are adapted, the modeling result can be directly used for finite element analysis, and the problems of complicated modeling, insufficient data and poor universality in the prior art are effectively solved.

The application belongs to the field of brain-computer interface, and particularly relates to a convolutional neural network construction method for enhancing time domain information in convolutional output, which is used for classification tasks or regression tasks in brain-computer interface, and comprises the following steps: improving and training the structure of a convolutional neural network to be enhanced, wherein the improvement mode is as follows: according to the structure of a convolutional layer to be improved, the size of a feature representation output by the convolutional layer is calculated, an average pooling layer with the same output size and feature representation size is constructed, the average pooling layer is used for accompanying convolution kernel sliding point multiplication process of the convolutional layer to be improved, input data is average-pooled to generate an average pooling representation; a trigonometric function encoder is constructed to perform time sequence coding on each element in the average pooling representation to obtain a new average pooling representation; a summation unit is constructed to sum corresponding elements of the new average pooling representation and the feature representation to obtain a feature representation with enhanced time domain information, and the feature representation is input into a next layer structure. The application can enhance time domain information without expanding the size of model parameters.