30 results about "Regression modelling" patented technology

The invention discloses a heavy-duty tool thermal deformation prediction method taking environmental temperature into account. The method specifically comprises the following steps of predicting the tool thermal deformation caused by the internal heat source of a tool, predicting tool thermal deformation caused by the external heat source of the tool, summing the tool thermal deformation caused by the internal and external heat sources to obtain the final tool thermal deformation. Time-lag thermal deformation error caused by the environmental temperature is taken into account during the prediction of the tool thermal deformation caused by the external heat source of the tool, and multiple regression modelling based on the least square method principle is performed during the prediction of the tool thermal deformation caused by the internal heat source of the tool. The joint action effects reflecting the influence of the environment temperature nonlinear lag and the influence of the internal heat source on the heavy-duty tool are combined by the method, and the thermal deformation error under any environmental condition and processing condition can be effectively predicted in real time.

The invention discloses a soft measurement method based on optimal selection and regression of orthogonal components in order to predict the orthogonal components conducive to quality indexes by meansof optimal selection and utilize the selected orthogonal components to create a soft measurement model for optimal regression. Specifically, according to the method, principle component analysis (PCA), independent component analysis (ICA) and an partial least squares algorithm (PLSR) are parallelly and separately utilized to obtain the corresponding orthogonal components at first, then adjacent component analysis based on a genetic algorithm is utilized to optimally select the orthogonal components, and finally, the selected orthogonal components are utilized to conduct optimal regression molding based on a particle swarm algorithm. Compared with a traditional method, the method has the advantages that characteristic components conducive to prediction of the quality indexes are selected in an optimal mode, and final predicted values of the quality indexes are obtained according to an optimal regression vector. Therefore, the soft measurement performance of the method is fully ensured,and the method is a more optimized soft measurement method.

The invention provides a combination optimization-based near-infrared non-invasive blood glucose detection wavelength variable screening method. The method comprises the following steps of firstly obtaining transmissivity spectral data of human blood glucose detection by adopting LED near-infrared light sources with different wavelengths; secondly selecting an optimal variable group in groups by using a continuous projection algorithm, a genetic algorithm and a gradual selection algorithm for multiple pieces of wavelength variable data by taking a root-mean-square error as an index in combination with linear regression modeling; thirdly performing scoring on each group of variables through a weighted scoring method; and finally recombining three groups of variables, sorting comprehensive scores in sequence from high to low, and selecting first few variables with highest scores as final auxiliary variables. According to an algorithm, a most effective wavelength combination is extracted through weighted combination optimization for the problem of an over-fitting phenomenon of a model due to excessive wavelength variables in regression modeling; and the variables are accurately selected, so that the model can be greatly simplified, the calculation complexity of the model is lowered, and the prediction capability and the real-timelines and robustness of non-invasive blood glucose detection are improved.

The invention belongs to the technical field of web application deployment and resource scheduling. In order to comprehensively consider the use conditions of different resources to carry out capacityexpansion and shrinkage to guarantee the application work efficiency, the web application-oriented container cluster self-adaptive capacity expansion and shrinkage system comprises a resource monitoring module, a regression modeling module and a container expansion and shrinkage module, and the resource monitoring module is used for obtaining the use conditions of container resources; the regression modeling module is used for extracting container load data, wherein CPU utilization, memory utilization, network throughput rate and file system throughput rate are characteristic values, the number of requests processed per second is a label, multiple linear regression modeling is carried out so as to obtain the influence of independent variables Ccpu, Cmemory, Cnetwork _ io and Cfs _ io on dependent variables Cqps, and therefore the known independent variables are used for predicting the change of the dependent variables. The method and the device are mainly applied to web application occasions.

The invention discloses a cross-border active user identification method and device. The method comprises the steps of obtaining behavior data of historical users on a cross-border APP; based on the behavior and score mapping relationship, determining scores corresponding to the behavior data, and based on the scores, determining cross-border activeness of historical users; randomly selecting a seed user from all historical users; obtaining multi-dimensional feature data of the seed user, taking the multi-dimensional feature data as a training data set, and taking the cross-border activity ofthe seed user as a label of the training data set; cleaning and vectorizing the training data set to obtain a processed training data set; performing regression modeling based on the lightGBM and theprocessed training data set to obtain an activity regression prediction model; and obtaining multi-dimensional feature data of other channel users, and predicting the activity of using the cross-border APP by the other channel users based on the activity regression prediction model. The active customer mining work efficiency can be improved.

The invention discloses a displacement filling modeling method and device, and the method comprises the steps: carrying out the regression modeling based on the relative displacement sc and relative speed vc of a plurality of groups of other vehicles and a target vehicle, which are obtained through the recognition of historical moment point cloud data corresponding to a plurality of other vehicles, and the corresponding calculation time tc, and obtaining a first displacement filling model; and determining a target displacement filling model based on the determinable coefficient R2 of the firstdisplacement filling model and a preset determinable coefficient value. Thus, the method is based on the relative displacement sc and the relative speed vc of a plurality of groups of other vehiclesand a target vehicle and corresponding calculation time tc. The target displacement filling model obtained by adopting the regression modeling mode is used for realizing displacement filling, and thepoint cloud data can more accurately predict the relative displacement of other vehicles and the target vehicle through the target displacement filling model, so that the delay difference is reduced,and the problem of inaccurate lag is avoided.

The invention relates to the technical field of financial science, in particular to a rejection inference method based on Cox regression and logistic regression and electronic equipment. The method comprises the following steps of: S1, collecting all application user data in a preset period, and defining two sets of tags for each information user, i.e., a binary classification target variable anda survival analysis target variable; S2, performing Cox regression modeling on useful information user data based on variables defined by survival analysis; S3, based on a Cox regression result, respectively calculating a probability P(G|A) that a rejected sample is a 'good' sample and a probability P(B|A) that the rejected sample is a 'bad' sample after credit granting and loan passing; and S4, based on the binary classification label and the rejected sample inference result, training a model by using a binary classification algorithm to complete the development of a score card model. According to the method, the survival analysis model is used for deducing a rejected sample, a non-parametric method Cox regression analysis is selected, a distribution function of the survival duration doesnot need to be judged, the process is simplified, deviation caused by unreasonable selection of the distribution function is avoided, and it is guaranteed that the model effect is more accurate.

The invention discloses a face key point detection method and system based on local principal component analysis. The method comprises the steps of S1, collecting a large amount of human face image sample data, and marking the face key points; S2, dividing the face key points into a plurality of local key points, adopting the principal component analysis to process the local key points, and obtaining the principal component features of the local key points; S3, calculating a combination coefficient of the key points of each face image under the principal component characteristics; S4, constructing a regression model, training the model through the combination coefficient, and generating a combination coefficient regression model; S5, inputting a to-be-detected face image into the combination coefficient regression model, and predicting to obtain a combination coefficient; and S6, restoring the face key points based on the predicted combination coefficient and the principal component features. According to the present invention, the local principal component analysis is carried out on the key points, and the local principal component coefficients are predicted, so that the complexity of directly carrying out principal component analysis on all the key points is reduced, and the regression modeling precision is improved.

The invention discloses an infrared spectrum modeling method based on a consensus selection technique. The infrared spectrum modeling method comprises the steps: according to original infrared spectroscopy data of samples, building a plurality of derivative spectrum spaces with different orders; in the plurality of derivative spectrum spaces with different orders, building respective calibration sets; treating the calibration sets of the derivative spectrum spaces by using the consensus selection technique, to obtain a basic calibration set; according to the basic calibration set, treating remaining samples in the derivative spectrum spaces with different orders, to obtain an extending calibration set; according to the basic calibration set and the extending calibration set, forming a final calibration set; and using the final calibration set and a validation set, and carrying out regression modeling. Through building the plurality of derivative spectrum spaces with different orders, then the derivative spectrum spaces with different orders are subjected to calibration set partitioning by using the consensus selection technique, the final calibration set formed from the basic calibration set and the extending calibration set is used for regression modeling, the model prediction accuracy is high, and the stability is good.

The invention discloses a manufacturing industry big data-oriented regression modeling method, which comprises the following steps: S1, obtaining low-dimensional features suitable for modeling through data preprocessing; s2, converting low-dimensional data of different service domains into a latent variable form; s3, establishing a regression equation among different latent variables through partial least square regression analysis, calculating to obtain the latent variables according to the maximum covariance among the latent variables, and determining the number of the latent variables by adopting a predicted residual sum of squares, so as to realize simultaneous regression analysis of multiple dependent variables on multiple independent variables; and S4, establishing a binomial regression equation between the latent variables to obtain a standard regression coefficient beta of each independent variable acting on each dependent variable, and further obtaining a single service predicted value. According to the invention, the latent structure model among the business domains is established, the influence relationship among different business domain data is mined, and different types of data of a plurality of business domains are communicated, so that the modeling effect of a single business is better, and the business is helped to improve quality and efficiency.

The invention discloses a natural gas demand prediction method and device, electronic equipment and a medium. The method can comprise the following steps: constructing a gray equal-dimension progressive complement model; according to a gray equal-dimension progressive complementation model, performing preliminary prediction on the influence factors to obtain predicted values of the influence factors; taking natural gas historical consumption data as dependent variables, taking historical values of influence factors as independent variables, and constructing a partial least square regression model; and predicting the natural gas demand according to the predicted values of the influence factors and a partial least square regression model. According to the method, the system noise is eliminated through the gray equal-dimension progressive complement model, and the development trend of various influence factors is predicted and analyzed; the partial least square model is used for carrying out regression modeling on the natural gas historical consumption and various influence factors, so that not only can the natural gas demand be reasonably and accurately predicted, but also the development trend and the change rule of the natural gas demand can be accurately mastered; the method has important guiding significance in natural gas production management, scheduling optimization, development planning, strategy making and the like.

The invention provides a blind regression modeling method for protecting data privacy in a mobile group perception system. Blind regression modeling is achieved by the interaction between a mobile perception node and a mobile perception server and can be summarized as the following steps: selecting a ''clean'' perception data subset, constructing a rough global model and performing global regression model refinement. The invention further provides an updating method of a blind regression model built by the blind regression modeling method. Model updating performed by using new perception datacan be summarized as the following steps: constructing a new rough global model and performing new regression model refinement. According to the method provided by the invention, an aggregation resultis exchanged between the mobile perception server and the mobile perception node to ensure that the contents of the perception data are not disclosed; and the incremental model updating is adopted toreduce the communication and calculation costs of the mobile perception node. By adoption of the method provided by the invention, the effects of protecting the perception data privacy, weakening theinfluence of abnormal data on the regression model, improving the model accuracy and realizing lightweight model updating are achieved.

The invention discloses a regression modeling method based on regression attention generative adversarial network data enhancement. A regression attention generative adversarial network adds attentionmechanisms to a generator and a discriminator; an attention module 1 in the generator constructs regression loss by utilizing independent variables and dependent variables of generated data output bythe generator; meanwhile, the attention module 1 is finely adjusted through real data; an attention module 2 in the discriminator constructs a new loss by using a difference value between the real data and the generated data regression loss; according to the invention, the feature containing the maximum regression information is extracted by minimizing the loss, and the feature contains the regression difference information between the maximized real data and the generated data, so that consideration of a discriminator on the regression information is facilitated. Based on the regression attention generative adversarial network, the original data is enhanced by utilizing the generated data, and then regression modeling is carried out by utilizing a data driving method, so that the performance and the prediction precision of the regression model are effectively improved.

Aiming at the problem that the evaluation of the project popularity has one-sidedness since the defect report and the feature report are separately explored for the existing research, the invention provides a project popularity analysis method based on a mixed effect linear regression model, comprising: by collecting project data from GitHub and then using statistical analysis and regression modeling, giving the influence relation on the project popularity by the number of defect reports and the number of characteristic reports in the project, analyzing the relation between the improvement ofthe project popularity and the defect reports and feature reports through the difference of the influence factors of the defect reports and feature reports in the project to the project popularity; and progressively performing four dimensions of analysis on the description diversity of the defect reports and feature reports to find out the difference between the defect reports and the feature reports in describing the diversity. The invention comprehensively studies the project popularity by analyzing the difference between the number of defect reports and the number of feature reports in theproject, and may comprehensively evaluate the popularity of the project.

The invention discloses a regression modeling method based on a Monte Carlo algorithm, and relates to the field of medicine and pharmacology, and the Monte Carlo algorithm is suitable for carrying out a calculation simulation test on a discrete system. In the calculation simulation, a probability model similar to the system performance is constructed, and a random test is performed on a digital computer, so that the random characteristics of the system can be simulated. Through the method, the efficiency of combination optimization can be improved.