87 results about "Multicollinearity" patented technology

The invention discloses a method for predicting ship fuel consumption according to sea conditions and manipulating conditions. The method for predicting the ship oil consumption comprises the steps of firstly acquiring a sample data set through data screening, data integration and normalized processing, then building a multivariate linear regression prediction model, defining a cost function by adopting a least shrinkage and selection operator (LASSO algorithm) based on the sample data set, performing variable shrinkage and parameter selection by combining cross validation and a least angle regression (LARS) algorithm, finally solving LASSO estimation by adopting an Osborne dual algorithm, and calculating to acquire the ship fuel consumption. The method provided by the invention for predicting the ship fuel consumption according to sea conditions and manipulating conditions can establish a function relation between the ship fuel consumption amount and each influence factor, solves a problem of multicollinearity in fuel consumption calculation, improves the ship fuel consumption calculation accuracy and has importance significance for energy conservation and emission reduction of traffic on the sea.

The invention relates to a modeling method of an urban electrical network distribution transform weight overload mid-term forewarning model. The forewarning model is established according to the following steps of firstly, collecting original data of correlated variables required for modeling, and cleaning the original data so that the quality of data entering the forewarning model can be ensured; secondly, designing and calculating characteristic variables of the forewarning model, screening the characteristic variables, and establishing a judgment basis for testing the multicollinearity among independent variables; thirdly, establishing one forewarning model on the basis of Logistic regression and through a stepwise regression method, and then judging whether the multicollinearity exists among the independent variables of the model or not so as to judge whether the model can be used or not; fourthly, repeatedly executing the second step and the third step so that the characteristic variables can be calculated again, and establishing various different forewarning models, evaluating the established forewarning models, and then comparing the evaluation parameters of all the forewarning models to determine the optimal forewarning model; fifthly, outputting the optimal forewarning model. By means of the method, the accurate distribution transform weight overload mid-term forewarning model can be easily established.

The invention belongs to the technical field of automobile industry data forecasting, and in particular relates to a method for forecasting electric vehicle ownership based on multiple linear regression and proportional substitution. Analyze the correlation with the target variable, establish and verify the multicollinearity model of traditional car ownership; according to the polynomial fitting results of the data of each influencing factor in known years, predict the unknown year data of each influencing factor, and substitute it into the above multiple After the collinear model, the traditional car ownership in the future years is predicted; combined with the local replacement ratio of electric vehicles and the actual growth of electric vehicles, new replacement increments are obtained, and then the total number of electric vehicles in the future years is predicted. The invention can use statistical data to calculate the number of traditional automobiles and calculate the number of electric vehicles, which is helpful for the planning of electric vehicle charging facilities and policy analysis.

The invention discloses a method for predicting forest fire hazard day occurrence probability , comprising steps of 1) determining fire hazard historic data, meteorological data, FWI factors of all groups, and Julian dates as impact factors having influence on the forest fire hazard and processing data, 2) performing multicollinearity diagnosis through data abnd choosing impact factors the expansion factors of which are smaller than 10 through, 3) choosing the impact factors having obvious impact on the fire hazard through Logistic Forward Wald analysis, and other two steps. The invention gives full consideration to the related factors having impact on the fire hazard in the region, performs evaluation on the probability of occurrence of forest fire hazard through a quantification method and has more accuracy on determination of dangerousness of occurrence of fire hazard.

The invention discloses a space-time continuous PM2.5 inversion method. The method comprises the steps of establishing a random forest regression model , wherein meteorological dynamic indexes of an inversion area and a satellite AOD serve as interpretation variables to be input into a random forest network, then training the random forest regression model to obtain an optimal model, and conducting inversion under the optimal model to obtain a PM2.5 concentration satellite estimation value; Determining a root-mean-square error of the PM2.5 concentration satellite estimation value; Calculatinga spatial interpolation observed by each station by using a common Kriging interpolation algorithm; Determining a root-mean-square error of the spatial interpolation; And using an inverse variance weighting method to fuse the PM2.5 concentration satellite estimation value and the spatial interpolation to obtain a final PM2.5 concentration inversion value. According to the method, the influence ofmultiple collinearity on an inversion result is overcome, the good noise resistance is achieved, seamless estimation of the near-surface PM2.5 concentration is achieved, and data support is provided for real-time continuous air quality monitoring of an area.

The invention discloses an aeromagnetic compensation method based on major constituent analysis. The method comprises steps that major constituents of a standard attitude matrix of calibration flight are calculated; the major constituents of the standard attitude matrix of the calibration flight are ordered according to contribution degrees, and multiple major constituents with relatively high contribution degrees are selected to acquire a new attitude matrix and a transformation matrix of the calibration flight; calibration flight data is utilized, the least square algorithm is utilized to acquire a compensation coefficient under the new attitude matrix; according to the transformation matrix of the calibration flight, major constituents of a standard attitude matrix of verification flight are extracted to acquire a new attitude matrix of the verification flight; magnetic compensation for measurement data of the verification flight is carried out, and aeromagnetic compensation based on major constituent analysis is realized. The method is advantaged in that problems of information overlapping caused by multicollinearity and inaccurate magnetic compensation caused by unstable inverse matrix solution existing in the prior art are effectively solved, the system information and noise can be effectively distinguished, system modeling accuracy is improved, and effective compensation of an aeromagnetic field is realized.

The invention discloses a method for estimating the needs of a transaction in a processor in a multi-tier architecture. The method comprises the following steps of: detecting inter-transaction correlation by analyzing an access log of a production environment; if strongly correlated transactions are discovered, extracting the shortest session subsequences and subsequences 1 shorter than the shortest session subsequence of the strongly correlated transactions by analyzing the access log; generating a logically correct artificial load of various transaction types by utilizing the subsequences; running the artificial load in a controllable testing environment to obtain performance monitoring data under the load; and finally obtaining the average servicing time of the transactions by using a conventional regression-based servicing time estimation method. By the method, the problem of multicollinearity of a conventional regression method in little load proportion change is solved, and the accurate average servicing time of various transaction types can be calculated and assist in making a rational capacity plan, thereby reducing the procurement cost of an enterprise and ensuring the performance of a network service.

The invention belongs to the technical field of traffic safety, and particularly relates to a traffic accident prediction method based on hybrid geographically weighted regression, which comprises thefollowing steps: step 1, dividing a spatial research area of a traffic accident, and collecting influence factor data; 2, explaining variables through multiple colinearity verification, and deletingunreasonable explaining variables; step 3, constructing a space weight function as a Gaussian function and a double square function; 4, determining that the bandwidth selection type is a fixed bandwidth and an adaptive bandwidth, and determining that a bandwidth optimization criterion is a corrected red pool information criterion; step 5, constructing and determining an optimal geographically weighted Poisson regression model; step 6, respectively bringing in explanatory variables as global variables to construct a hybrid geographically weighted Poisson regression model to perform a comparisontest; and step 7, constructing and determining an optimal hybrid geographically weighted Poisson regression model. The invention provides a traffic accident prediction method based on hybrid geographically weighted regression, which is sufficient in spatial heterogeneity consideration and high in prediction model precision.

The present invention provides a method and device for forecasting order quantity. The method includes establishing a decision table; processing the decision table by using the rough set theory based on the database system to obtain a reduced attribute set; using the reduced attribute set to establish a regression analysis model; and using the regression analysis model to predict the order quantity. The method, device and system solve the problem that the variance of the estimator of the regression coefficient becomes larger due to multicollinearity, and realize a more accurate prediction of the order quantity.

The invention discloses a multi-model space-time modeling method based on a finite Gaussian mixture model. The method is applied to a nonlinear distribution parameter system, a nonlinear space obtained by the nonlinear distribution parameter system is divided into a plurality of local operation subspaces based on a finite Gaussian mixture model, and an original complex nonlinear space-time dynamicequation is summarized into a plurality of simple nonlinear space-time dynamic equations, so that local modeling is carried out; when all the local space-time models are integrated, a principal component regression method is adopted to calculate the weight of each local space-time model, the existence of multiple colinearity is avoided, and a global space-time model of a large working area is reconstructed through multi-model modeling. The method provided by the invention has better performance for a large-scale, strong-nonlinearity and time-varying system.

The invention provides a credit scoring method, a system, a computer device and a readable medium. In a data acquisition stage, multi-channel information, such as operator information, credit card information and debit card information, is selected, and various aspects of information of a user are comprehensively considered, so that the prediction effect and stability of the model can be enhanced.Derived fields based on time slices can extract the information with real prediction value, and also can increase the prediction effect of the model. In the data preparation phase, the variables withhigh missing rate and the variables with difficult-to-interpret level are deleted, which enhances the stability of the whole model. In the model development phase, the variables with multicollinearity are deleted, which enhances the stability of the model. LASSO and other machine learning methods are used to select the variables with real prediction ability, which improves the prediction abilityof the final model.

The present invention provides a forest biomass-based remote sensing image feature selection method and a forest biomass-based remote sensing image feature selection apparatus. The method includes the following steps that: feature values are extracted from a forest remote sensing image, the feature values are preprocessed through an SR (stepwise regression) algorithm, and feature values corresponding to multicollinearity are removed from the preprocessed feature values, so that a feature set can be generated, the initial set of the feature set is a full set; the feature set is updated repeatedly according to the following processes: an SVM (support vector machine) algorithm is trained according to the initialization feature set, so that the weights of feature values in the initialization feature set are determined, an SVM-REF (support vector machine-recursive feature elimination) algorithm and the weights are adopted to construct the feature sequencing coefficient of the feature values in the feature set, and the feature values in the feature set are sequenced according to the feature sequencing coefficient, and the feature set is updated according to the sequence of the feature set; and update operation is carried out continuously until the number of feature values in the current feature set is equal to a preset number of feature values, and the current feature set is determined as the optimal feature set used for forest biomass. With the method and apparatus of the invention adopted, the effect of remote sensing image feature selection can be optimized.

The invention provides a multivariate robust soft measurement method about sewage treatment effluent quality index, and relates to the technical field of the sewage treatment automatic control. The method comprises the following steps: taking a parameter real-time measured by conventional detection equipment based on industrial field as input data of a model; establishing a random weight neural network model capable of simultaneously performing multivariate dynamic prediction on the main parameters for balancing the sewage treatment effluent quality, thereby realizing robust soft measurement of BOD content, COD content, TSS content sewage quality parameters, and comprehensively describing the sewage quality parameters, thereby avoiding the hysteresis of the offline assay and the uncertainty of the manual operation. The sparse partial least square and Schweppe type general M are utilized at the same time so as to eliminate the influence on the modelling by multiple colinear and reduce the bad influence on the modelling by outlier and leverage point in the data; and meanwhile, the aim of variable selection is reached, and an estimation value of the multivariate sewage treatment effluent quality parameter at the specified dynamic time zone can be given more accurately.

The invention provides an intrauterine insemination success rate influence factor calculating method for a polycystic ovarian syndrome patient based on logistic regression and a system thereof. The method comprises the steps of collecting structured information of a case, extracting pregnant information and a related characteristic; converting the related characteristic so that the characteristiccan be accepted by a user and an algorithm; training a logistic regression model by means of an intersected verifying manner; outputting a model parameter which corresponds with each kind on the condition of an optimal super-parameter, performing highest and lowest normalization on all model parameters, and restraining the parameter to an interval of [0,1] as an influence degree output. Compared with traditional chi-square testing methods, the method according to the invention has advantages of realizing more reasonable hypothesis, reducing influence caused by multiple colinearity of the variable, and realizing good interpretability.

The invention discloses an improved LR-Bagging algorithm based on feature selection, which comprises the following steps: first, determine the initial data set from the original data, and require that the degree of correlation between the independent variable and the dependent variable cannot be too low; secondly, the initial data set Discrete independent variables are encoded by WEO; then random sampling is used to obtain a certain number of records and feature fields to form training examples, and the training examples are trained with LR ((LogisticRegression) model and the normal significance test of the coefficient is performed. If not significant, then Eliminate, on the contrary, add the combination model. Carry out cyclic iterations until the combination model is better. Finally, you can use the better combination model for prediction and grouping. This algorithm can improve the diversity of classification results, the degree of extraction of variable information and the prediction results It can also effectively reduce the possibility of multicollinearity and "overfitting" caused by too many variables in the base LR model.

The invention discloses a software failure time forecasting method based on a kernel partial least squares regression algorithm. Through the application of a kernel function technology, the problem of software reliability forecast is converted to the problem of recession estimation, and the kernel partial least squares regression algorithm is used for resolving the problem of the software reliability forecast. Through fully consideration of a small sample property of the software reliability forecast, the situations that the size of observational variables is bigger than that of observational samples and multicollinearity exists among the variables can be overcome by using the kernel function technology, and so that a model 'overfitting' situation arises in modeling approaches such as a neural network does not occur. By means of the software failure time forecasting method based on the kernel partial least squares regression algorithm, model parameters are automatically and continuously adjusted to fit the dynamic change in a failure process, therefore adaptive forecasting of the software reliability is achieved, and the adaptive capability of a software failure forecasting model is improved effectively.

The invention discloses an electric vehicle charging safety early warning method. The method analyzes the consistency, performance degradation and safety characteristic coupling mechanism of a battery pack of an electric vehicle, and deduces the SOC consistency development rule of the battery pack; in order to further play the battery performance to the greatest extent and guarantee the safety of the battery pack, the safety state monitoring method of the battery pack is studied; in order to solve the multicollinearity problem in the power battery pack, the capacity of the lithium ion battery is estimated through principal component regression PCR, then regression is conducted on data through a multiple linear regression method, the result is compared with the result of the health state estimation model based on principal component regression, and finally, for short-term potential safety hazards, an outlier detection method is adopted to quickly and accurately detect an abnormal battery and give an early warning, and for long-term potential safety hazards, a power battery medium-and-long-term accelerated decline identification method is adopted to give an early warning. The method provides reference for safe charging of the electric vehicle.

The invention discloses a sample component measurement method based on a PCR-ELM algorithm. The sample component measurement method comprises the following steps: S1, acquiring infrared spectrum data of a sample to be measured; S2, obtaining the contents of the respective components in the sample corresponding to the infrared spectrum data according to a PCR-ELM model. According to the sample component measurement method, the PCR-ELM model is used for processing the infrared spectrum data of the sample to be measured to obtain the contents of the respective components in the sample corresponding to the infrared spectrum data. Compared with the method adopting a conventional model for data processing, the method disclosed by the invention has the advantages that the phenomenon of overfitting is avoided, and the multicollinearity among variables is reduced; furthermore, the fitting precision is improved, the forecasting precision of the spectrum data with a small sample amount and high dimension, and the stability of the forecasting precision are improved, and the application range of the ELM algorithm is expanded.

A method of determining an optimal statistical model that can best show the statistical characteristics of given data and an apparatus performing the method are provided. The method acquires target data to be analyzed, where the target data consists of a plurality of independent variables and a dependent variable. Then, the method determines one or more independent variables based on variances in the target data, establishes a first statistical model that shows the relationship between the m independent variables and the dependent variable, and calculates first error of the first statistical model. The method generates a plurality of first statistical models by repeatedly performing the steps of establishing the first statistical model and calculating the first error while changing the value of m, and selects a statistical model with minimum error as an optimal statistical model for the target data. In this manner, a statistical model having the multi-collinearity between independent variables minimized and having an improved precision can be selected.

The invention discloses a research method based on a GAMLSS model sediment transport contribution rate. The method comprises the steps of firstly collecting and arranging flow and sediment informationof a drainage basin outlet station; then computing a relation between each index and sediment transport and a fitting process line, computing 90% P-factor and R-factor, 50% quantile sequence varianceand a mean value, a correlation coefficient and an AIC; then using an attribution analysis method to analyze six indexes, thus acquiring an influence degree of each index on sediment transport variation; and at last, computing a contribution rate of each index to sediment transport. According to the research method based on the GAMLSS model sediment transport contribution rate provided by the invention, the problem that the prior art cannot accurately and comprehensively acquire the influence degree of the climate change and the human activities on the sediment transport contribution rate issolved, the principal component regression analysis is used, the influence of multicollinearity is eliminated, the influence generated by each variable mean is considered, and the influence of each variable interannual variance change on sediment transport is also considered.

The invention discloses a method for predicting daily occurrence probability of lightning fire. The method includes: building a data file, and taking each grid point and all fire points in a testing area as sampling points; performing multi-collinearity diagnose through data; selecting factors evidently influencing fire occurrence through Logistic Forward Wald analysis; building a daily occurrence probability model of the lightning fire through a Logistic model so as to analyze daily occurrence probability of the lightning fire, determining a lightning fire ignition threshold through a secondary judging theory, and calculating precision so as to perform precise analysis; on the basis of GIS, substituting day value data into the model for calculation so as to complete forecast of the daily occurrence probability of the lightning fire, and the like. The method has the advantages that the problem that the existing lightning fire predicting models rely on a lightning monitoring network is solved, wide application is achieved, the method can be used for calculating the daily occurrence probability of the lightning fire and analyzing future lightning fire occurrence trend, and the existing lightening fire pre-warning models are supplemented effectively.

The invention relates to a CPP-based hydrological model parameter dynamic calibration method. The method comprises the following steps: S1, periodically dividing the rate into a plurality of sub-periods on an annual scale, and calculating meteorological clustering indexes and underlying surface clustering indexes of all the sub-periods; screening out candidate clustering indexes based on a nonlinear relationship between the clustering indexes and the flow; S2, eliminating multiple colinearity among clustering indexes by adopting a PCA algorithm; S3, averaging the clustering index values of each sub-period of all years; performing clustering operation twice in sequence according to the meteorological indexes and the underlying surface indexes, and finally dividing the hydrological process in the year into four sub-periods for calibration; S4, independently optimizing TOPMODEL model parameters in each sub-period by adopting an improved parallel calibration scheme, and combining the TOPMODEL model parameters to generate continuous flow sequence values; and S5, evaluating the simulation performance of the hydrological model under different flow conditions by using a multi-index comprehensive evaluation system. According to the invention, the prediction capability of the hydrological model in a changing environment is effectively improved.