33 results about "Multicollinearity" patented technology

The invention discloses a spare part demand prediction method based on a ridge regression improved algorithm, and relates to the technical field of spare part demand prediction, and the method comprises the following steps: 1, data collection and preprocessing; 2, performing feature extraction on the time series data; step 3, constructing an improved ridge regression algorithm: combining Huber loss and a kernel function; 4, dividing a training set and a test set according to a time sequence, selecting an optimal parameter by adopting a K-fold cross validation method, performing robust kernel ridge regression model training on the training set, predicting a model generated by training on the test set, comparing with a true value of the training set, and calculating MAE, MSE and a value; and step 5, comparing with a traditional ridge regression prediction result. According to the method, the model overfitting problem caused by multiple collinearity among influence factors in spare part demand prediction can be solved, compared with traditional ridge regression, the nonlinear relation among variables can be captured, robustness is higher, and therefore the spare part demand quantity can be predicted more accurately.

The invention discloses a method for analyzing influence of climate change and human activity on space-time evolution of water resources. The method comprises the following steps: acquiring meteorological information, hydrological information, land utilization information, soil information, social economic information and water resource information of a to-be-detected area; performing trend analysis on the water resource information to obtain a space-time evolution rule of the water resource information; performing attribution analysis on the time-space evolution rule to obtain influence factors of the time-space evolution of the water resource; obtaining the correlation degree of the influence factors and the space-time evolution rule, and completing the analysis of the climate change and human activity on the space-time evolution of the water resource. According to the method, the influence factors of water resource evolution of the to-be-detected area are researched through the multivariate statistical stepwise regression analysis model, the multicollinearity between the related influence factors of climate change and human activity is reduced through the characteristics of stepwise regression analysis, the calculation accuracy is improved, and the sustainable development of water resources is promoted.

The invention discloses a multi-classification risk prediction method and system for liver diseases based on machine learning, and relates to the technical field of biomedicine, and the method comprises the following steps: collecting fatty liver disease diagnosis results and biochemical indexes of a subject to form a training sample set, the method comprises the following steps: screening out biochemical indexes significantly related to fatty liver diseases through single-factor regression analysis, determining potential risk factors, carrying out multicollinearity test on the factors, screening out risk factors, constructing a plurality of machine learning classification models for training, and selecting a model with the best performance as a reference model. The contribution degree of each important risk factor is evaluated and sorted, classification significant factors are determined, the significant factors serve as classification metadata, a plurality of judgment models are trained, input factors are dynamically selected according to contribution values, and finally a fatty liver disease degree classification result is output, so that complex conditions under different sample features and clinical backgrounds are better handled; and the accuracy of prediction results is improved.

The invention discloses a segmented quantitative analysis method for uplift pressure of a gravity dam. The uplift pressure of the gravity dam is influenced by various factors, and safety monitoring data of the gravity dam often have the problems of multiple collinearity and small samples. The method comprises the following steps: introducing a BFAST time sequence decomposition method, decomposing original uplift pressure monitoring data into a trend component and a seasonal component, and obtaining trend segment points and seasonal segment points; a Bayesian model average modeling method integrating Bayesian model selection is introduced, and a gravity dam uplift pressure statistical model is established in a segmented mode; and finally, upstream and downstream water pressure components of the uplift pressure are separated according to the average multiple models, and quantitative analysis of the uplift pressure of the gravity dam is achieved. According to the method, more accurate fitting and component separation effects can be obtained, and accurate quantitative analysis of the uplift pressure of the gravity dam can be realized.

The present disclosure relates to a method and device for determining influencing factors of flue gas emissions, equipment and medium. A plurality of initial influencing factors and a plurality of flue gas emission information of a target unit are obtained; the initial influencing factors and the flue gas emission information are respectively subjected to time sequence causality analysis, dependency analysis and correlation analysis to obtain time sequence causality information, dependency information and correlation between the initial influencing factors and the flue gas emission information; then, based on the time sequence causality information, the dependency information and the correlation, candidate influencing factors are obtained from the plurality of initial influencing factors, so that the selected influencing factors have high confidence in causality, explainability and information integrity; finally, the system feature dimensionality of the candidate influencing factors is reduced to determine the target influencing factors that affect the generation of a plurality of flue gas emission information of the target unit, thereby eliminating the multicollinearity between different influencing factors. Thus, the explainability of the pollutant emissions of the unit under the complex working condition of deep peak regulation is realized.

The application discloses a kind of based on elastic network regression's wiring error leakage user positioning method, system, terminal and medium, wherein method includes: obtaining the abnormal station area leakage fault day's station area residual current data, user load current data, and constructs station area residual current time series and subordinate user load current time series;With the user load current data obtained as explanatory variable, station area residual current data as explained variable, carry out elastic network regression calculation, obtain the optimal explanatory variable after eliminating multicollinearity and its corresponding regression coefficient and construct regression model;Compare the absolute value of each regression coefficient, and the user whose absolute value of regression coefficient is greater than preset threshold is judged as zero line, ground line wiring error user.Through identifying zero line, ground line wiring error abnormal user, to solve the existing low-voltage station area exists because user zero line, ground line wiring error and lead to user load current data into residual current problem.

This invention provides a method and related equipment for assessing earthquake landslide susceptibility. Multiple landslide influencing factors are selected from multi-source data of the study area, and multicollinearity analysis and Pearson correlation coefficient analysis are performed on all landslide influencing factors to obtain basic environmental factors. Newmark displacement is calculated based on the physical and mechanical parameters in the multi-source data and the basic environmental factors. The Newmark displacement is stacked with the basic environmental factors to obtain multi-channel image data, which is then input into an earthquake landslide susceptibility assessment model for evaluation, resulting in a coseismic landslide susceptibility zoning map of the study area. Compared with existing technologies, this invention uses Newmark displacement as an independent feature input, compensating for the lack of physical mechanism support in traditional pure data-driven models. It achieves the complementary advantages of geological disaster dynamics mechanisms and deep learning feature extraction capabilities, improving the accuracy and reliability of earthquake landslide susceptibility assessment.

The application discloses a software performance anomaly detection method and device. The software performance anomaly detection method comprises the following steps: collecting data of attributes representing the running state of a target software, and obtaining a first data set; screening out elements not meeting a preset multicollinearity condition in the first data set according to a multicollinearity test standard, and obtaining a second data set; performing noise reduction processing on the second data set by using an adaptive discrete wavelet decomposition method according to the waveform characteristics of key performance elements, and obtaining a third data set; training an anomaly detection model by using the third data set, and obtaining a software performance anomaly detection model; and detecting the target software to be detected by using the software performance anomaly detection model, and obtaining a detection result. By combining the multicollinearity test standard, the adaptive discrete wavelet decomposition method and the anomaly detection model, the problem of excessive noise and unobvious software performance failure characteristics is effectively solved, and the accuracy of software detection is improved.

The present application is suitable for the field of mineral exploration, and provides a lithium geochemical anomaly identification method and system for a lithology complex area, which comprises the following steps: obtaining geochemical data of the area, and preprocessing the geochemical data of the area; determining PLSR independent variable indexes according to the preprocessed geochemical data; constructing a PLSR regression model according to the PLSR independent variable indexes; determining lithium geochemical background values of each sample point in the area according to the PLSR regression model; and identifying lithium geochemical anomalies in the area according to the predicted upper limit of the lithium geochemical background values. The present application uses PLSR to construct a regression model between lithium and lithology indicating elements, and then can determine the lithium geochemical background values of each sample point, effectively solves the multicollinearity problem in lithium multiple linear regression analysis, improves the calculation accuracy of the lithium geochemical background values of each sample point in the lithology complex area, and lays a solid foundation for lithium geochemical anomaly identification.

The invention belongs to the technical field of big data analysis, and discloses an immersed tunnel cost prediction method and system based on big data analysis. According to the method, firstly, an immersed tunnel cost big database is constructed, and key feature parameters are screened out through correlation analysis and multi-collinearity processing; then, on the basis of engineering attributes of the cost influence factors, layering the feature vectors, and respectively establishing an intra-layer nonlinear mapping relationship and an inter-layer coupling relationship reflecting a synergistic effect between different engineering attribute feature layers, so as to construct a cost prediction model fusing multi-level influence; and finally, correcting the preliminary prediction result in combination with the specific parameters of the project to be predicted and the human-material-machine price dynamic data to obtain the final cost. According to the method, the construction cost of the immersed tube tunnel can be accurately and efficiently predicted.

This invention discloses a method for predicting the risk of landslides based on connectivity and a geographic detector. It employs a discretization approach to transform the comprehensive soil erosion index and influencing factors into characteristic parameters of categorical variables. Multicollinearity analysis is performed on all characteristic parameters to filter out those exhibiting collinearity. The geographic detector is used to calculate the explanatory power of each characteristic parameter for landslide risk in the target area and the frequency of landslide occurrence at sample points within different grids. The product of the explanatory power and frequency of landslide occurrence for different characteristic parameters within each grid is accumulated to obtain a spatial distribution map of landslide risk prediction for the target area. The beneficial effects of this invention are: by combining soil erosion and sediment transport potential, and using a geographic detector to analyze the explanatory power of each characteristic parameter for landslide occurrence, highly accurate landslide risk assessment results are obtained.

The invention discloses an air conditioner load prediction method based on PCA and GRU. The method comprises the steps that firstly, historical air conditioner load data and corresponding temperature, weather and date type multivariate influence factors are obtained; carrying out dimension reduction on the multivariate influence factors by utilizing a principal component analysis method, eliminating multiple collinearity, and extracting a plurality of comprehensive principal components of which the cumulative contribution rates exceed a preset threshold value; carrying out abnormal value identification, correction and normalization preprocessing on the historical load data; aligning the preprocessed load data with the comprehensive principal component according to a time sequence, and constructing an input feature sequence for training a gating cycle unit model; and finally, predicting the air conditioner load in a future time period by using the trained model. According to the method, model input is effectively simplified through PCA, the data quality is improved, the strong time sequence modeling capability of the GRU model is combined, the precision and efficiency of short-term air conditioner load prediction are jointly improved, and reliable technical support is provided for power system dispatching and demand side management.

The invention discloses a typical tree species growth prediction method, system and device based on multi-stage multi-factor regression and a medium, and relates to the technical field of tree species growth prediction.The method comprises the steps that multi-source heterogeneous data are collected and preprocessed; dividing the independent variables into different types of influence factors, and performing statistical test on each factor to obtain a preliminary candidate variable set; calculating a variance expansion factor of the candidate variables, and when the expansion factor exceeds a threshold value, reducing the correlation among the preliminary candidate variables by adopting a collaborative path method to obtain a final variable; performing regression modeling through a three-stage modeling method based on the final variable to generate a regression model, and establishing a regression sub-model for each partition; and based on the obtaining mode of the final variable, extracting a judgment rule of the tree species and outputting the judgment rule in a structured format. According to the method, high-precision prediction of the growth under multi-factor driving can be realized, and the problems of multi-collinearity, unstable variable selection, insufficient nonlinear structure expression and the like in a traditional regression model are solved.

This application relates to the fields of remote sensing image processing and ecological environment monitoring technology, and discloses a method for extracting peat bog information based on the dense temporal characteristics of multi-source remote sensing. The method includes acquiring time series of microwave radar and optical vegetation index, performing alignment and cleaning, extracting low-frequency and continuous first-order derivative sequences from the microwave radar to generate temporal basis features; extracting historical delay features for effective optical observation times, splicing the basic state term and derivative modulation product term to generate a bilinear distributed lag observation matrix; performing scale unification processing on the lag observation matrix, constructing a diagonal regularized matrix based on time-lapse attenuation characteristics, and solving for the solution vector; calculating the normalized asymmetric hysteresis intensity and reconstruction quality factor, and combining dual thresholds to determine the output distribution data. This invention incorporates the physical hysteresis effect of hydrological evolution into the computational framework, overcoming multicollinearity and observation sparsity problems, and achieving high-precision extraction of peat bogs.

The application provides a balanced intrusion detection method and system based on width focus learning, and belongs to the field of information security intrusion detection. The method comprises the following steps: obtaining a standardized data set, screening and filtering information features of the standardized data set through information variance and multicollinearity, eliminating invalid features and redundant features, and optimizing data quality; adopting a generative adversarial network to construct a balanced data set, introducing a deep belief network to optimize generated data, increasing data diversity and authenticity, and using a sigmoid model as a discriminator of the adversarial network. Finally, a width learning network is used to learn and train the balanced data set, a focal loss mechanism is introduced to strengthen the attention degree of the width learning model to different types of attack samples, so that the detection and recognition of different attack samples are achieved. The balanced intrusion detection model based on width focus learning improves the precision and reliability of intrusion detection attack detection, shortens the training time, and improves the detection rate of different attack samples.

The invention discloses a karst region soil thickness mapping method based on a stacked integrated model, and relates to the technical field of soil mapping, and the method comprises the following steps: taking climate, terrain and stony desertification elements as spatial information; obtaining optical remote sensing images of the research area in recent ten years, processing the remote sensing images to obtain vegetation, soil and stony desertification indexes, and generating various topographic factors through topographic software; performing correlation analysis on environment covariables to eliminate multiple collinearity, converting qualitative variables and unifying resolution; dividing the data into two groups containing and not containing stony desertification information, and eliminating and screening an optimal variable by using six models through recursive features; taking the six models as base models, and taking prediction results as input to construct three meta-models to realize stacking integration; and finally, soil thickness prediction mapping is completed, and the uncertainty is predicted through quantile regression quantification. According to the method, the stacked integrated model and stony desertification information are combined, and the precision and resolution of soil thickness prediction are improved.

The invention discloses a method for predicting low-temperature performance of aged asphalt, and belongs to the technical field of performance evaluation of road materials. According to the method, firstly, an asphalt sample is subjected to simulated aging treatment, then improved column chromatography is adopted for separating and measuring the content of saturates, aromatics, colloids and asphaltene of aged asphalt, a chromatographic column is of an activated aluminum oxide and silica gel double-layer adsorption structure, gradient elution is combined, and the separation efficiency and reproducibility are remarkably improved. The method comprises the following steps: firstly, measuring a creep rate m value of aged asphalt, analyzing the correlation degree between the m value and components of the aged asphalt by using grey correlation, and establishing a multiple linear prediction model based on ridge regression by taking the contents of four components as independent variables and the m value as a dependent variable; the model effectively overcomes the multicollinearity problem among components, component analysis data are input into the model for unknown samples, the low-temperature performance of the samples can be predicted, and the research and development efficiency and the engineering quality control level are greatly improved.

The invention discloses a stepwise regression analysis method-based college student physique test score prediction model, and belongs to the technical field of sports and metering economics. The method comprises the following steps: collecting and preprocessing college student physique test multi-dimensional index data; constructing a multiple linear regression model between the total physical test score and a plurality of explanatory variables; screening significant variables by adopting a stepwise regression analysis method, and optimizing a model structure; performing multi-collinearity, heteroscedasticity and sequence correlation test and correction on the model; and finally verifying the validity of the variable set by using a random forest model. According to the method, key influence factors can be automatically identified from numerous physical indexes, a prediction model with high goodness of fit and high interpretation is constructed, accurate prediction of student physical scores is realized, and a scientific basis is provided for physical health management of colleges and universities.

The application discloses an electric energy metering error monitoring method, a computer readable storage medium and a processor, relates to the technical field of distribution network automation systems, and solves the problem that mutual information regularization imposes equal punishment on all similar users, leading to error estimation collapsing to an average value. The application divides user historical data into sections according to power consumption, respectively estimates section estimation errors in each section, and then calculates individualized coefficients representing sensitivity. On this basis, the scheme constructs weight coefficients with the sensitivity as a key adjustment factor, thereby protecting the independence of error estimation when solving the objective function and avoiding being forcibly pulled towards the average value of other similar users. In summary, while maintaining mutual information regularization to overcome multicollinearity and improve estimation stability, the application effectively preserves the differences in metering error characteristics of individual users, significantly improving the identification accuracy of out-of-tolerance meters in complex power consumption scenarios.

A method for constructing a predictive model for the 28-day mortality risk of patients with severe traumatic brain injury (sTBI) includes: collecting clinical data of patients undergoing sTBI surgery; screening samples that meet the inclusion and exclusion criteria and grouping them according to 28-day survival; collecting clinical indicators of the samples and calculating serum sodium variability on the 8th postoperative day; screening candidate variables using LASSO regression and simplifying variables through correlation analysis and multicollinearity diagnosis; dividing the dataset into training and test sets, and determining independent predictors through univariate and multivariate logistic regression analysis; constructing a nomogram prediction model based on the independent predictors, and validating the model performance through ROC curve, calibration curve, and decision curve analysis. The predictive model of this invention uses GCS score, oxygenation index, and 8-day serum sodium variability as core indicators, exhibiting high predictive accuracy and strong clinical operability, providing a quantitative basis for prognostic assessment and clinical intervention for sTBI patients.

The invention discloses an epidural delivery analgesic lower head dystocia risk prediction method and system. The method comprises the following steps: retrospectively collecting clinical feature data of a parturient to be parturified under epidural delivery analgesia; the method comprises the following steps: determining classification tangency points of continuous variables according to clinical medicine consensus, literature review or statistical distribution, converting the continuous variables into ordered or disordered classification variables, performing code conversion on the collected classification variables, and constructing a structured data set; carrying out preliminary screening on all feature data by using single-factor logistic regression analysis and carrying out multi-collinearity diagnosis; the screened features are incorporated into a multi-factor logistic regression model for optimization to determine key prediction variables, and a logistic regression prediction model is constructed to derive a logistic regression equation; and converting the logistic regression model into a Nomogram column graph model, and outputting a prediction result of the occurrence risk probability of the head dystocia. According to the scheme, early risk prediction of epidural delivery analgesia lower head dystocia is realized, and an earlier decision window is provided for clinic.

Techniques are disclosed as an optimization data system for eliminating correlated independent variables programmatically from data with ranked exclusion scores. The system can obtain an initial dataset comprising variables, determine a set of correlation values by analyzing linear correlation between the variables, generate a correlation matrix using at least in part the set of correlation values and corresponding variables from the initial data, calculate exclusion scores for the variables in the correlation matrix that exhibit multicollinearity, and update the initial dataset by removing at least one variable with the highest exclusion score from the variables to generate an updated dataset comprising optimized variables. The steps for correlation and elimination of variables are iterated until an updated dataset without any correlation is obtained and then a machine learning model may be trained using the updated dataset.

This invention discloses a method for constructing a risk prediction model for blinding diabetic retinopathy (STDR), belonging to the field of diabetic lesion detection. It addresses the problem that STDR screening at the grassroots level relies on specialized resources and that existing models have poor adaptability. The method includes: screening eligible type 2 diabetic patients and organizing clinical data; detecting 15 core laboratory indicators and calculating derived indicators as candidate variables; dividing patients into non-STDR and STDR groups according to DR classification and DME diagnosis results; selecting variables using a dual-dimensional strategy of "statistical significance + clinical relevance" through binary logistic regression, multicollinearity test, and incorporating six indicators including age, duration of diabetes, and MLR to construct a mathematical prediction model; and then building a nomogram visualization prediction model based on the model formula. All indicators included in this model are routinely available in clinical practice. The model has been validated with an AUC of 0.825, sensitivity of 87.2%, specificity of 67.8%, good fit, and good clinical applicability. It can quickly assess the risk of STDR and is suitable for grassroots medical scenarios.

This invention relates to a method for predicting the performance of hydrogen storage alloys based on feature engineering and multi-model screening, comprising the following steps: constructing a set of physicochemical features of the alloy; multi-method feature importance analysis; feature union extraction; feature screening based on importance-correlation joint analysis; feature subset search and multi-model cross-validation; and determination of the globally optimal model and the optimal feature subset. This invention provides a more comprehensive approach to feature construction and selection for subsequent machine learning research on hydrogen storage alloys; it can effectively identify features that are important in different methods, making the feature selection results more objective and reliable. Simultaneously, it eliminates multicollinearity among features through correlation redundancy removal and avoids the risk of local optima through full subset search, significantly improving the stability, reliability, and global optimality of feature screening; the prediction accuracy of this invention is improved by 90.8% and 55.9%, respectively.

The application provides a kind of urban road collapse safety risk intelligent diagnosis method, belongs to the technical field of urban lifeline engineering safety, comprising: determining the target area road network range, obtaining the detailed catalog of road collapse in target area and disaster factor data;Build road collapse disaster factor set, and carry out data preprocessing and multiple collinearity analysis;Build road collapse sample data set, divide it into training set and test set, and carry out data enhancement on training set;Using extreme gradient boosting algorithm to capture the nonlinear relationship between disaster factors, build urban road collapse prediction classification model;Based on the classification model, the probability of road collapse is predicted, and the road collapse risk distribution map of the target area is prepared;The constructed classification model is analyzed for interpretability, and the importance of each disaster factor and its effect on road collapse is evaluated.

The invention discloses a global climate long-sequence data reconstruction method for heat stress risk assessment, and the method comprises the steps: carrying out the spatial interpolation of a meteorological data missing value of a target station through a ridge regression model based on the data of a spatially adjacent reference station; and time interpolation is carried out through a ridge regression model based on the physical relationship among different meteorological variables of the target station. According to the method, a ridge regression technology is adopted, coefficient estimation is stabilized by introducing a regularization term, multiple collinearity is effectively overcome, space-time collaborative interpolation is achieved, and the method is high in precision and self-adaptive.

The application utilizes variational mode decomposition (VMD) and principal component analysis (PCA), and proposes a VMD-PCA-based sub-synchronous oscillation energy characteristic influence factor identification method. The port energy characteristics of sub-synchronous oscillation under different working conditions are extracted through time domain simulation, and a regression model is established to perform regression analysis on the port energy of sub-synchronous oscillation. The process is as follows: the VMD mode decomposition is performed on the measured voltage and current at the outlet of the fan to obtain the voltage and current components under the sub-synchronous oscillation mode, and the transient energy flow at the outlet of the fan is obtained by calculation; the transient energy flow function of the fan outlet is fitted, and the energy flow power is taken as the stability characteristic quantity of sub-synchronous oscillation for research; the principal component method is used to solve the multicollinearity problem between the energy characteristic influence factors, a variable fitting evaluation model of the energy characteristic is established based on the stepwise regression method, and the key influence factors of the sub-synchronous oscillation energy characteristic are identified. The method can judge the current system energy characteristic of the system in time, identify the key influence factors affecting the system energy characteristic, and has important engineering significance for maintaining the safe and stable operation of the new power system.

The application relates to a multi-condition frequency coupling impedance identification method and device using multiple linear regression, wherein the method comprises the following steps: obtaining a frequency coupling admittance matrix of a measured device under multiple conditions to obtain a condition parameter variable matrix; based on the condition parameter variable matrix, a permutation importance index of the condition parameter variable is calculated, a first condition parameter variable set satisfying a first preset key condition of the measured device is determined; based on the first condition parameter variable, the multiple collinearity degree of the condition parameter variable is calculated, a second condition parameter variable set satisfying a second preset key condition is obtained, and a multiple linear regression model for identifying the multi-condition frequency coupling impedance of the measured device is constructed. Therefore, the problems in the prior art that when the number of hidden layers of the neural network impedance fitting algorithm exceeds the model generalization requirement, the training set error and the test set error will produce significant deviation, and when extrapolated to unknown conditions, the error will significantly increase and the generalization ability will be insufficient are solved.

The invention is suitable for the field of mineral exploration, and provides a lithium geochemical anomaly identification method and system for a lithologic complex region, and the method comprises the following steps: obtaining the geochemical data of the region, and carrying out the preprocessing of the geochemical data of the region; according to the preprocessed geochemical data, determining a PLSR independent variable index; according to the PLSR independent variable indexes, constructing a PLSR regression model; according to the PLSR regression model, determining a lithium geochemical background value of each sample point in the region; and identifying lithium geochemical anomalies of the region according to the prediction upper limit of the lithium geochemical background value. According to the method, the regression model between the lithium and the lithology indication elements is constructed by utilizing the PLSR, so that the lithium geochemical background value of each sample point can be determined, the multi-collinearity problem in lithium multiple linear regression analysis is effectively solved, the calculation accuracy of the lithium geochemical background value of each sample point in a lithology complex region is improved, and the lithium geochemical background value of each sample point in the lithology complex region is calculated. And a solid foundation is laid for lithium geochemical anomaly identification.

The invention provides a prediction method for thermal stability of a coordination polymerization organic metal catalyst. The method comprises the following steps: constructing a model of a catalyst active species, an ethylene monomer, a catalyst ligand and a bimolecular ethylene coordinated complex, carrying out geometric structure optimization and frequency analysis to obtain a thermodynamically stable 3D structure, and collecting and sorting corresponding stable molecular structures and thermodynamic correction values of enthalpy values and Gibbs free energy of the stable molecular structures; obtaining corresponding single-point energy, enthalpy value and Gibbs free energy of each molecular structure, bond dissociation enthalpy and bond dissociation energy; quantum chemical descriptors are calculated; combining the quantum chemical descriptors, the key dissociation enthalpy and the key dissociation energy to construct a database; and adopting multiple linear stepwise regression analysis to remove multiple collinearity of the descriptors, and constructing a prediction model. According to the prediction method provided by the invention, the thermal stability of the coordination polymerization organic metal catalyst can be accurately predicted, so that the workload is reduced, and efficient screening of the polymerization catalyst is realized.