70 results about "Simple linear regression" patented technology

The invention belongs to the field of lithium ion batteries, and discloses an online SOH estimation method of a lithium ion battery for solving the problems that characteristic parameters are difficult to be obtained online, the dependency of a model on training data is high, the required data size is large, the complex function relationship between the battery capacity and the characteristic parameters is difficult to be described by simple linear regression, and the estimation accuracy is difficult to be guaranteed in an implementation process of the existing SOH estimation technology. According to the online SOH estimation method disclosed by the invention, the characteristic parameters are obtained from a capacity increment curve by using a capacity increment method. The method does not require the battery to undergo a complete charging and discharging process, the feature parameter extraction is simpler, and the application of the method in the BMS is facilitated. The establishment of a characteristic parameter and SOH function model is completed by using a multi-output Gaussian process regression model method, the potential correlation between different outputs is better used, and the estimation accuracy of SOH is improved. Meanwhile, the dependency of the method on the training data is small, and the online SOH estimation method has very good adaptability on different types of lithium ion batteries.

A method of calibrating glucose monitor data includes collecting the glucose monitor data over a period of time at predetermined intervals, obtaining reference glucose values from a reference source that temporally correspond with the glucose monitor data obtained at the predetermined intervals, calculating the calibration characteristics using the reference glucose values and corresponding glucose monitor data to regress the obtained glucose monitor data, and calibrating the obtained glucose monitor data using the calibration characteristics. In additional embodiments, calculation of the calibration characteristics includes linear regression and, in particular embodiments, least squares linear regression. Alternatively, calculation of the calibration characteristics includes non-linear regression. Data integrity may be verified and the data may be filtered.

The invention relates to a mobile video quality assessment method based on hierarchy analysis and multiple linear regressions. The method comprises the following steps: firstly, determining each layer of end-to-end performance indexes of user QoE (Quality of Experience) influencing mobile video services; dividing various end-to-end performance indexes of the QoE into a target layer, a criterion layer and an index layer by utilizing a hierarchy analysis method according to the attributes and the types of the performance indexes; respectively constructing hierarchical influence models in manners that the index layer is aligned to the criterion layer which is aligned to the target layer from bottom to top by utilizing the multiple linear regression analysis; constantly adjusting regression coefficients for influencing the models and various performance indexes of the models; and finally, establishing the influence model in the manner that the index layer is aligned to the target layer, i.e., establishing a total regression model influencing the quality of a mobile video according to various end-to-end performance indexes of QoE so as to assess the quality of the mobile video by utilizing the hierarchical total assessment model. According to the method, the subjective and objective performance indexes of the mobile video are combined with the QoE of the user, so that the method provided by the invention is a comprehensive and effective user experience quality assessment method.

The invention discloses a method for analyzing the ice and climate relationship on the basis of the grey MLR (Multiple Linear Regression), which comprises the following steps of: a step 1 of acquiring microclimate data comprising the environment humidity, the environment wind speed and the environment temperature and wire tensile force by a power transmission line online monitoring system and converting the microclimate data into an actual ice thickness value by a theories calculation model according to the wire tensile force; a step 2 of selecting effective data and applying a grey GM (1, 1) model construction to forecast the microclimate data; and a step 3 of establishing an ice thickness model by applying a MLR analysis method according to the selected effective data, forecasting the ice thickness on the basis of the microclimate GM (1,1) model construction and establishing an ice thickness forecasting model of the grey multiple regression analysis. The invention solves the problems that in the prior art, an equation system shows morbidness due to accumulation and subtraction of the data, the parameter estimation shows instability and the like. Detailed information such as the weight variation before and after a monitored power transmission line is iced, the inclination angle of an insulator string, the environment temperature and humidity, the wind speed and the like can be provided; and the ice thickness of the power transmission line is obtained by calculation according to field acquisition information.

The invention discloses a remote sensing estimation method for the SPAD value of rape leaves in different growth periods. The method comprises the following steps: by taking commercial crop rapes in northwest regions as research objects, measuring the spectral reflectivity and SPAD value of rape leaves in each growth period; analyzing the relativity between 10 types of spectral indexes and the SPAD value of the rape leaves; constructing an SPAD random forest regression estimation model of the rape leaves based on the spectral indexes, and verifying the constructed model; and comparing the constructed model with a traditional simple linear regression model and a traditional multiple stepwise regression model. Modeling and verifying R2 respectively reach 0.91 and 0.74 or above, the verifying RMSE is between 1.571 and 5.004, RE is between 2.66% and 13.22%, and thus an optimal estimation model of the SPAD of the rape leaves can be obtained.

A method for building a fine grain tailing project property index estimation empirical formula based on linear regression includes the following steps: 1) collecting and settling fine grain tailing data physico-mechanical index sample data; 2) adopting the mu+ / -3sigma abnormal value rejection principle to conduct preliminary screening on the sample data; 3) conducting secondary screening on the sample data based on mathematical statistics; 4) building a linear regression mathematical model; 5) determining model parameters; 6) conducting model goodness inspection; 7) building the fine grain tailing project property index estimation empirical formula. The method has the advantages of providing forceful support for scientificity and reliability of a data regression analysis result through massive test data information, providing reliable parameter choice for fine grain tailing base stability evaluation and design in future, greatly saving project investment cost and soil engineering test investment, avoiding unnecessary project building cost, enabling a research result to have popularization and application value and being simple, practical, reliable in result, efficient in computing and the like.

A city center evaluation method based on mobile phone signaling comprises three steps of pretreatment, recognition and evaluation. According to OD records, two kinds of city resident behaviors (one is commuting and the other one is life going out) are recognized, and commuting and life going out OD information serves as basis data for recognizing and evaluating two city centers; simple linear regression is adopted, as for a unit j around each target grid unit i in an R range, the distance Dij between the unit j and the unit i is calculated, the visit density value yj at the j is extracted, and univariate linear regression is carried out on the yj and the Dij. As for selection on a proper radius R, with a greedy algorithm as reference, a point which needs to be recognized as a center is determined in advance, searching of the radius R is expanded from a small radius step by step, one recognition result is generated in the case of one time of radius searching, and the best predetermined center recognition effect is selected from the results as the final recognition result.

The invention relates to a building energy consumption modeling method based on linear regression. The method comprises the steps that historical data of all energy consumption units in a building is obtained; all the historical data is preprocessed; a regression model is built through classification data by applying linear regression analysis fitting; a future expected energy consumption value of an energy consumption part and a future expected total energy consumption value of the building are predicted according to an objective function. According to the building energy consumption modeling method, the consumption condition of building energy can be effectively predicted, by controlling the influence condition, the energy conservation and consumption reduction purposes are achieved.

The invention provides a lithium battery life prediction method based on degradation path coordinate reconstruction and multiple linear regression. Firstly, a new sampling rule is used for carrying out coordinate reconstruction on degradation data of a newly-developed battery and truncated similar batteries; then, the mapping relation between the life of the similar batteries and the life of the newly-developed battery is established through reconstruction data; finally, the life of the newly-developed battery is calculated through a service life prediction model. According to the method, the service life of the newly-developed battery can be predicted with the help of full-life degradation data of the similar batteries under different formulations or structures or use environments, the test time of a development stage life test can be greatly shortened, the test amount of the development stage life test can be greatly reduced, the development period of new products can be shortened, the cost at the development stage can be reduced, and core competitiveness of enterprises can be improved.

The invention relates to a multivariate linear regression fire risk assessment method based on big data, which is used in the field of data processing and analysis. Comprises the following steps of: (1) constructing a hierarchical structure model of fire risk factors; (2) dividing a building fire risk grade; (3) Establish the evaluation model R=WTX; (4) Constructing the data sample of risk assessment machine learning; 5, calculate that weight W according to the sample, and establish a fire safety fire risk assessment model R=WTX; (6) Calculating the attribute value of the building to be evaluated according to the step (4.1), and comparing the risk value calculated by the model established in the step (5) with the risk grade divided in the step (2) to obtain the risk grade of the building.This method adopts multi-attribute evaluation method, realizes the quantitative evaluation of fire risk grade, and comprehensively evaluates the fire safety management risk of social units.

The invention discloses a method for achieving EMS load prediction based on a decision tree and linear regression. According to the method, the characteristics of the high algorithm classification accuracy of the decision tree, the simple decision tree generation mode and the simple linear regression interpolation logic are sufficiently utilized, energy consumption values which are recorded on time at certain energy consumption nodes every day and historical information formed with schedule days (from Monday to Sunday) and temperature which are highly related to energy consumption are added into the system, and the best time span of historical data adopted for the decision tree is obtained by using the natural selection method through the historical data. The method has the advantage of better meeting the requirement for prediction of short, medium and long term loads of an energy consumption unit.

A method and associated systems for using machine-learning methods to perform linear regression on a DNA-computing platform. One or more processors generate and initialize beta coefficients of a system of linear equations. These initial values are encoded into nucleobase chains that are then padded to a standard length. The chains are allowed to bind with complementary template chains in a DNA-computing reaction, and the resulting DNA molecules are decoded to reveal the relative the relative likelihood of each chain to bind. The initial values of the beta coefficients are weighted proportionally to these likelihoods, and the process is repeated iteratively until the beta coefficients converge to optimal values.

The invention discloses a multivariate linear regression algorithm applied to cost analysis, belonging to the technical field of cost analysis, comprising the following steps: establishing a database,subdividing, sorting and classifying the cost of raw materials, electricity and gas, consumption of office supplies, wages of workers, cost of fixed assets consumption, etc.; The data in the databaseare substituted into the multiple linear regression equation, and the basis of data analysis is established. Establish a description of data sources, both external and internal; Through the model calculation and analysis process and model results, the model results include observation value, index value, prediction value, residual and standard residual; The method includes regression statistics,variance analysis and regression parameter table. The invention uses multiple linear regression as an analysis tool to study the relationship between each main cost and the total cost, and reasonablyanalyze and predict the total cost according to the variation trend between the consumption amount of each unit cost and the total cost.

The invention discloses an image reversible information hiding method and apparatus based on multiple linear regression. The method comprises the following steps: selecting a plurality of pixel pointswithin a neighborhood of a target pixel point as prediction samples according to a certain rule, and selecting a plurality of pixel points as training samples in the neighborhood of each prediction sample to serve as training samples; establishing a multiple linear regression function relationship between the training sample pixels in a local area and the prediction sample pixels; and predictinga target pixel value by using the function relationship between the training sample pixels and the prediction sample pixels. According to the image reversible information hiding method and apparatus,the accurate prediction of the target pixel value is achieved by using the internal connection between the pixels around the target pixel, and the performance of the image reversible information hiding and embedding is further enhanced.

The present invention relates to an experimental apparatus for predicting the multi-point vibration response frequency domain under the condition of the unknown load; an experimental data generation method for predicting the multi-point vibration response frequency domain under the condition of the unknown load; and a method for predicting frequency domain vibration response of the unknown measure point according to the frequency domain vibration response of the known measure point by using the experimental apparatus and the experimental data, and by using the multiple linear regression model and the least squares generalized inverse method of the linear relationship between frequency domain response data under the unrelated multi-source unknown load combined excitation. The multiple linear regression model and the least squares generalized inverse method of the linear relationship between frequency domain response data are directly used instead of knowing or identifying the transfer function, the load size, or even the load position of the system. According to the technical scheme of the present invention, mainly for the environment of the unrelated multi-source unknown load combined excitation, vibration response prediction of the unknown node is carried out by using the vibration response prediction of the known measure point, so that vibration response situation of one unknown node and a plurality of unknown nodes can be predicted.

Methods, systems, and computing system program products for iteratively calculating Simple Linear Regression (SLR) coefficients for streamed data, including iteratively calculating one or more components of SLR coefficients for an adjusted computation window based on one or more components of SLR coefficients calculated for a pre-adjusted computation window and then calculating the SLR coefficients for the adjusted computation window based on the iteratively calculated components. Iteratively calculating SLR coefficients avoids visiting all data elements in the adjusted computation window and performing redundant computations thereby increasing calculation efficiency, saving computing resources and reducing computing system's power consumption.

The invention provides a locally weighted linear regression based ultra-dense network load balancing optimization method which jointly regulates cost offset values of all small stations. The locally weighted linear regression based ultra-dense network load balancing optimization method comprises the steps of firstly fitting daily load data collected by a base station by utilizing a locally weighted linear regression method to obtain a load curve of the base station; providing a relatively optimal iterative initial value for a cost based distributed user connection method; and solving the load balancing problem in an ultra-dense heterogeneous network. As a logarithmic function is adopted as a utility function, compromise of opportunity and fairness of resource distribution among users is realized, and users at the edge and in the middle of the base station realize 3.5 times and 2 times data throughput gains respectively. The cost value of each base station is updated through the distributed iteration, and the loads of the base stations in different layers and in the same layer are balanced automatically, and low-complexity load balancing is realized. By setting the initial value through the locally weighted linear regression method and predicting the number of users accessed to the base station at one moment, the time of iterations and the computation complexity are reduced greatly.

The invention disclose a method utilizing two texture indexes to jointly determine the tuna flesh freshness, wherein a corresponding relationship between a K value and the tuna flesh freshness has already been known, and the two texture indexes are a hardness index and an elasticity index. The method also comprises the following steps: (1) obtaining a simple linear regression equation between the hardness index and the K value and a simple linear regression equation between the elasticity index and the K value, thus the relationship between the hardness index and the tuna flesh freshness and the relationship between the elasticity index and the tuna flesh freshness are obtained according to the corresponding relationship between the K value and the tuna flesh freshness; (2) unfreezing the freezed tuna flesh to be detected, cutting into slices after the unfreezing process, and storing the tuna flesh slices at a measured temperature; (3) carrying out a hardness index detection and an elasticity index detection on the tuna flesh sample to be detected; (4) comparing the hardness index and the elasticity index obtained in the step (3) with the conclusion obtained in the step (1) so as to obtain a freshness grade related with the hardness index and a freshness grade related with the elasticity index.

The invention relates to a high-speed maneuvering target tracking method based on linear regression and cluster analysis theories. The method is mainly suitable for radar to track a target which has a high speed and a wide-range variation of an accelerated speed, and the method can achieve stable tracking of various maneuvering conditions under the condition that real-time performances are guaranteed. The method includes implement processes that firstly, a sample set is established for historical parameters of a maneuvering target track, a self-adaption relevance wave door is designed according to regression analysis theories, and the track is split to form a plurality of tracks; then a sample with concentrated main tracks is projected to extract feature vectors, and a cluster center is acquired and the similarity in a cluster is calculated according to the cluster analysis theories; projection values of split track parameters are sequentially obtained by means of the same method, and the projection values are samples to be resolved; and finally, an interest target track is extracted according to a similarity discrimination criterion. According to practical engineering application of the method, the method is capable of achieving stable tracking of the high-speed strong maneuvering target and has wide application prospects with continuous improvement of maneuvering capability of future weaponry.

The invention discloses a semi-supervised image clustering subspace learning algorithm based on local linear regression. Firstly, a local linear regression model is used for predicting a coordinate of a training sample in a clustering subspace, a local prediction error between a predicted value and a true value is obtained, and then a minimized objective function of a total predicted error is obtained; then according to two constrain conditions of inter-class dispersion maximization and inner-class dispersion minimization, and a marked sample and an unmarked sample are used for calculating an inter-class dispersion matrix and a total dispersion matrix; and finally, the inter-class dispersion matrix and the total dispersion matrix are blended in the minimized objective function of the total predicted error to obtain an objective function for solving clustering subspace, and function solving is performed through generalized characteristic root to obtain the optimal clustering subspace. The semi-supervised image clustering subspace learning algorithm based on the local linear regression makes full use of the marked sample, the unmarked sample and a local adjacent relation in a training data set to obtain good clustering results.

Aiming at an HEVC video coding standard, the invention provides a high-definition video code rate control algorithm based on the Sobel operator and linear regression. The method mainly comprises the steps: finding an optimal initial QP (quantization parameter) for a coded first frame through a learned linear regression model, thereby enabling a coding process to quickly adapt to a set bandwidth; generating the intra-frame and inter-frame first-order gradient information through employing the Sobel operator, and then adaptively selecting optimal gradient information to replace coding complexity to serve as the weight of LCU layer bit distribution. The method enables the bit rate error to be lower, enables the output code rate to be more stable, and is better in subjective and objective quality of a video. An experiment result indicates that the method, compared with an HEVC standard code rate control scheme K0103, reduces the bit rate error by about 0.67%, and improves the peak SNR by about 0.44dB.

The invention relates to a battery state of health (SOH) estimation method for a general discharge working condition, which belongs to the technical field of batteries, and comprises the following steps: S1, directly performing voltage segmentation on a constant current or pulse working condition to obtain an electric quantity difference sequence; for a dynamic current working condition, carryingout voltage filtering firstly, then carrying out voltage segmentation, and obtaining an electric quantity difference sequence; S2, extracting health factors: calculating standard deviations of the electric quantity sequence and the electric quantity difference sequence as the health factors, and respectively representing the health factors as stdQ _ VP and stdQ _ VP; S3, analyzing the linear correlation between the health factor and the battery capacity by adopting a Pearson correlation coefficient; and S4, estimating the SOH of the battery by using a data driving method. According to the method, a simple linear regression model can be established to estimate the SOH of the battery under a universal discharge working condition.

The invention discloses a deformable mirror iteration control method and system based on multi-element linear regression. The system is a closed-loop system composed of a computer, a deformable mirror, an attenuator and an interferometer. The method comprises the step of solving control parameters of a target surface shape, wherein the step of solving the control parameters of the target surface shape comprises the sub-steps that a plurality of sets of different random control parameters are applied to the deformable mirror, corresponding Zernike coefficient vectors are recorded, and then an influence function matrix G of the deformable mirror is solved on the basis that A equals to G*V; a set of initial iteration control parameters are applied to the deformable mirror and a correspondingactual surface shape is measured; the difference between the actual surface shape and the target surface shape is calculated, and whether the difference is smaller than a preset threshold or not is judged; if yes, the initialized iteration control parameters serve as the control parameters of the target surface shape; if not, a surface shape difference control parameter is solved on the basis thatA equals to G*V to update the iteration control parameters, iteration continues till the surface shape difference is smaller than the preset threshold.

The present approach relates, in some aspects, to a multi-level and a multi-channel frame work for segmentation using model-based or “shallow” classification (i.e. learning processes such as linear regression, clustering, support vector machines, and so forth) followed by deep learning. This framework starts with a very low resolution version of the multi-channel data and constructs an shallow classifier with simple features to generate a coarser level tissue mask that in turn is used to crop patches from the high-resolution volume. The cropped volume is then processed using the trained convolution network to perform a deep learning based segmentation within the slices.

The invention discloses a spatial attribute prediction method based on local weighting linear regression. The method comprises the steps that a multivariate linear regression method is adopted for conducting mathematical modeling on an influence relation of spatial attributes, a spatial distance factor is introduced to a fitting assessment expression of linear regression analysis to solve the problem, and the spatial distance and the influence degree on an analysis target is in an inverse correlation, wherein data of low spatial distance has large influence on the analysis target, and data ofhigh spatial distance has less influence on the analysis target. The spatial attribute prediction method is low in computation complexity, unique and simple in calculation result and beneficial to program processing, the linear combination relation of various spatial attributes can be processed, and the problem of less fitting during combination analysis of multi-dimensional spatial attributes byadopting an existing linear regression analysis method is effectively solved.

The embodiments of the invention provide a multipath suppression method and a device based on linear regression. The method comprises the following steps: a processor collects samples corresponding to advance correlators and lag correlators at a target sampling time; the processor determines the slope at both sides of an autocorrelation function curve for the samples obtained according to a linear regression approach; the processor calculates the phase discrimination error based on the slope at both sides of the autocorrelation function curve and a target output value; and finally, the processor corrects the phase discrimination value of a phase discriminator based on the calculated phase discrimination error so that the phase discriminator can adjust the original local copy code of a positioning receiver according to the corrected phase discrimination value to complete multipath suppression, wherein the samples include the output values of the correlators at the target sampling time and the code phase deviation values corresponding to the correlators. Through the scheme of the invention, the accuracy of the calculated slope at both sides of the autocorrelation function curve can be ensured, the accuracy of the corrected phase discrimination value can be ensured, and the effect of multipath suppression can be improved.

A color carrier frequency offset estimation algorithm based on phase unwrapping and linear regression is disclosed, (1) providing the initial phase of color synchronized component of initial black level signals in a filed; adopting the black level signals at an initial place in the filed; the phase range obtained by the phase estimation being from 0 to 2Pi, there being five phases as frequency estimation; (2) using a phase unwrapping algorithm to unwrap the extracted phase; the phase unwrapping is that the phase at other four lines is adjusted in an unique linear range based on the phase at the first line; (3) using a linear regression model to fit the unwrapped phase, thereby extracting color synchronous frequency; determining the solutions of overdetermined equations by means of least square criterion; (4) using a limitr to limit the extracted color synchronous frequency. When the input is non-standard video signal, the color carrier frequency is estimated by means of the phase unwrapping algorithm and the linear regression model so as to improve the precision of the local color carrier and carry out digital high-definition demodulation.

Provided is a traffic congestion prediction method including: a step for detecting the acceleration of a vehicle; a step for calculating a power spectrum corresponding to a frequency from frequency analysis of the detected acceleration; a step for calculating the simple linear regression of the power spectrum and calculating the maximum value of the change in the gradient of the simple linear regression in a specified frequency range as the maximum gradient value; a step for detecting the inter-vehicle distance between the aforementioned vehicle and a vehicle in front; a step which uses a distribution estimation method to estimate an inter-vehicle distance distribution from the detected inter-vehicle distance; a step for calculating the minimum covariance value from the estimated inter-vehicle distance distribution; a step for estimating a front vehicle group distribution from the correlation relationship between the minimum covariance value and the maximum gradient value; and a step for predicting in real time traffic congestion based on the vehicle group distribution.

The invention discloses a multiple linear regression associative memory model-based fingerprint and face coupling recognition method. The method includes the following steps that: S1, a fingerprint picture and a face picture are collected; S2, the associative memory input matrixes and output matrixes of the fingerprint picture and face picture are obtained; S3, a multiple linear regression fingerprint picture recognition model with regression parameters and a multiple linear regression face picture recognition model with regression parameters are constructed; S4, the regression parameters are calculated, and the multiple linear regression fingerprint picture recognition model and the multiple linear regression face picture recognition model are obtained; and S5, the fingerprint picture and face picture are recognized. The multiple linear regression associative memory model-based fingerprint and face coupling recognition method can realize multiple recognition of identity information and has high reliability. According to the method, the associative memory and the multiple linear regression models are combined together, the pictures are converted into the parameters, and therefore, the method has the advantages of high safety factor, good recognition effect, good protection effect of the identity information and high confidentiality.

The invention provides a linear regression-based pedestrian recognition method. The method comprises the following steps: selecting two types of training image sets; processing the two types of selected training images; modeling for the two types of processed training images; performing multi-scale zooming on an image to be tested; selecting a candidate region through sliding window processing; processing the candidate region; next estimating a least square parameter in the processed data; then calculating a most approximate vector of the estimated numerical value; calculating the distance between an observation vector and the approximate vector; judging the type of the candidate region; filtering off a duplicate detection part by using a non-maximum suppression algorithm; finally outputting a recognition result. According to the method, no complex characteristic extraction and classifier training process is required, a simple structure is realized, and the real-time property of the recognition method is effectively improved.