CN120853757A - Microplastic prediction method based on fusion of feature engineering and meta-learning - Google Patents
Microplastic prediction method based on fusion of feature engineering and meta-learningInfo
- Publication number
- CN120853757A CN120853757A CN202511011100.9A CN202511011100A CN120853757A CN 120853757 A CN120853757 A CN 120853757A CN 202511011100 A CN202511011100 A CN 202511011100A CN 120853757 A CN120853757 A CN 120853757A
- Authority
- CN
- China
- Prior art keywords
- micro
- plastic
- model
- meta
- abundance
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a micro-plastic abundance detection base model based on feature engineering and meta-learning fusion, which is characterized by comprising the steps of constructing a micro-plastic abundance detection base model, inputting micro-plastic abundance detection features into the micro-plastic abundance detection base model, calculating to obtain a micro-plastic abundance prediction result, obtaining a corresponding micro-plastic abundance prediction random forest model and a micro-plastic abundance prediction gradient lifting model, constructing a meta-learning device, fusing the micro-plastic abundance prediction random forest model and the micro-plastic abundance prediction gradient lifting model through the meta-learning device to obtain new micro-plastic abundance detection features, inverting through cross verification to obtain the prediction result of the micro-plastic abundance detection base model, generating micro-plastic abundance detection meta-model, and carrying out iterative training on the micro-plastic abundance detection meta-model by utilizing the micro-plastic detection meta-model to obtain an optimized prediction result. The invention realizes the method for inverting the abundance of the seawater micro-plastic by using the fused machine learning model.
Description
Technical Field
The invention relates to the technical field of machine learning and the technical field of marine environment regression prediction, in particular to a regression prediction method of marine micro-plastic integrated by combining feature engineering, random forests, gradient lifting trees and meta-learner.
Background
At present, remote sensing inversion technology for microplastic is still immature, and cases are few. In the prior art, a single machine learning model (such as a random forest or gradient lifting tree) has the following problems that 1, the characteristic expression capability is insufficient when high-dimensional and nonlinear data are processed, the traditional model depends on original characteristics, and complex interaction relations among the characteristics are difficult to capture. 2. The model generalization capability is limited, and the prediction performance of a single model is easily influenced by super parameter selection and is sensitive to noise. 3. The meta-feature is not utilized enough, namely the correlation among the prediction results of different models is not fully mined by the existing method, and the precision is difficult to further improve.
For example, some prediction methods based on random forests in the prior art can process high-dimensional data, but do not introduce feature interaction terms, so that modeling of nonlinear relations is insufficient. And, some in the prior art, gradient lifting trees are adopted for prediction, but the box division strategy is fixed, the training set distribution is not dynamically adapted, and the deviation is easy to generate on the test set. Currently existing techniques for inverting the abundance of microplastic by remote sensing generally use a single model, which is susceptible to noise. The existing method does not fully mine the relevance between the prediction results of different models.
How to improve the fusion strategy of feature engineering and models is a technical problem to be solved in the invention.
Disclosure of Invention
The invention aims to solve the problem of insufficient characteristic expression capability of the prior art, and provides a microplastic prediction method and a microplastic prediction system based on fusion of characteristic engineering and meta-learning, which are used for capturing the interaction relation among characteristics through characteristic interaction by combining a random forest and a gradient lifting model so as to realize microplastic prediction for further improving prediction precision.
The invention is realized by the following technical scheme:
the invention discloses a microplastic prediction method based on fusion of feature engineering and meta-learning, which comprises the following steps:
s1, acquiring micro-plastic detection data, selecting and combining through characteristic engineering, and extracting micro-plastic detection characteristics;
S2, constructing a micro-plastic abundance detection base model, inputting the micro-plastic detection characteristics into the micro-plastic abundance detection base model, and calculating to obtain a micro-plastic abundance prediction result;
s3, training the micro plastic abundance detection base model through a random forest algorithm and a gradient lifting algorithm respectively and independently to obtain a corresponding micro plastic abundance prediction random forest model and a micro plastic abundance prediction gradient lifting model;
s4, constructing a meta learner, fusing the micro plastic abundance prediction random forest model and the micro plastic abundance prediction gradient lifting model through the meta learner to obtain new micro plastic abundance detection characteristics, and inverting through cross verification to obtain a prediction result of the micro plastic abundance detection base model to generate micro plastic detection meta characteristics;
And S5, constructing a micro-plastic abundance detection meta-model, and performing iterative training on the micro-plastic abundance detection meta-model by utilizing the micro-plastic detection meta-features to obtain an optimized prediction result.
In some embodiments, S1 further comprises:
Generating interaction items for the micro plastic detection characteristics, and performing interval variable transformation, wherein the interaction items generated according to the micro plastic detection characteristic matrix are respectively (B6+B7), B6, (B5+B6), B7, (B5+B7) and (B5+B7)/B4, B4-B7 are 400nm-2200nm wave band data of wave band remote sensing data from environmental monitoring, B4 is a wave band of 545 nm-560 nm, B5 is a wave band of 1230 nm-1250 nm, B6 is a wave band of 1628-1652 nm, B7 is a wave band of 2105 nm-2135 nm, and an interaction column of characteristic 1X characteristic 2 is generated and is expanded to 7 dimensions;
And dynamically dividing the interval type variable into boxes, calculating an optimal box dividing boundary, selecting the first column of characteristics of the training set, dividing the first column of characteristics into five equal-width intervals, dividing the numerical values in the training set into corresponding boxes, and storing the box dividing boundary.
In some embodiments, S3 further comprises:
constructing a random forest model, training a microplastic detection original sample data set model through the random forest model, evaluating the importance of the microplastic detection input data characteristics, and selecting the microplastic detection input data characteristics as optimized microplastic detection input data characteristics according to the importance;
Constructing a gradient lifting model, training a micro-plastic detection original sample data set through the gradient lifting model, training a learner by using a least square criterion, and making a decision target to minimize the sum of squares of residual errors of a micro-plastic abundance detection predicted result and a real result.
In some embodiments, obtaining the predicted result of the micro-plastic abundance detection base model and its derivative features through cross-validation in S4, generating micro-plastic abundance detection meta-features further comprises:
The micro-plastic detection input data trained by the random forest model and the gradient lifting model are randomly and uniformly divided into 5 subsets, 5-fold cross validation is carried out, and a meta-feature matrix [ RF_ oof, GB_ oof, RF_ oof ×GB_ oof and RF_ oof 2,GB_oof2 ] is constructed.
In some embodiments, S5 further comprises constructing a micro-plastic abundance detection metamodel:
and (3) standardizing meta-characteristics, and training the micro-plastic abundance detection meta-model, wherein the aim is that the sum of squares of abundance prediction results and real result residuals of micro-plastic detection is minimum, so as to obtain a micro-plastic abundance prediction value.
In some embodiments, S3 further comprises constructing a meta learner:
generating a random forest model for detecting the abundance of the microplastic and a gradient lifting model for detecting the abundance of the microplastic through 5-fold cross validation, carrying out meta-feature standardization as meta-features, expanding dimensions of the meta-features including an original predicted value, an interactive item RF_ oof ×GB_ oof, a square item RF_ oof2 and a square item GB_ oof2, carrying out meta-model training of the feature engineering by adopting a LSBoost algorithm, combining a decision tree-based learner, and iteratively optimizing a predicted result.
In some embodiments, the method further includes visualizing a scatter plot of the predicted results of the training set and the test set.
In some embodiments, the method further comprises performing a meta learner evaluation:
Reversely restoring the predicted results of the normalized training set and the test set into an original range;
And calculating a determination coefficient R 2 of an abundance predicted value and a true value of the micro-plastic detection, square root RMSE of a mean square error, mean absolute error MAE and mean square error MSE to evaluate the micro-plastic abundance detection metamodel.
Compared with the prior art, the invention can achieve the following beneficial technical effects:
1) The meta learner further integrates the advantages of multiple models, and experiments show that the testing set R 2 is improved by 10% -15%, so that the precision can be improved.
2) By reducing the number of training samples, the R 2 and RMSE of the model are reviewed. The training samples are better than the traditional model when being 84-40, 40-16 and smaller than 16, so that the model is better in stability compared with the traditional model, and the stability can be improved.
3) The method effectively reduces the risk of overfitting through grid search optimization parameters and a cross verification strategy, achieves the effect of robustness enhancement, specifically, adds noise to a training set of a model, reduces R 2 from 0.95 to 0.92, reduces the amplitude by about 3.2%, and increases RMSE from 0.63 to 0.67 per m3 by about 6.3%. Whereas the R 2 drop of the conventional model was 13.9% and the RMSE rise was 26.1%. The robustness of the model is thus enhanced.
Drawings
FIG. 1 is a flow chart of the overall method for predicting microplastic based on fusion of feature engineering and meta-learning;
FIG. 2 is a detailed view of the overall flow of the microplastic prediction method based on the fusion of feature engineering and meta-learning;
FIG. 3 is a technical roadmap of the microplastic prediction method based on the fusion of feature engineering and meta-learning of the invention.
Detailed Description
The following detailed description of specific embodiments of the invention will be given with reference to the accompanying drawings.
FIG. 1 shows a microplastic prediction method based on fusion of feature engineering and meta-learning, which comprises the following specific steps:
step 1, carrying out data preprocessing to obtain a microplastic detection original sample data set, wherein the original sample uses band remote sensing data about 400-2200nm band data, 545-565nm band data and 1230-1250nm environmental monitoring to invert the abundance of the seawater microplastic;
Step 1.1, cleaning data, and deleting samples containing missing values or abnormal values, such as data exceeding 3 times of standard deviation;
step 1.2, carrying out normalization processing, and normalizing input and output data to a [0,1] interval through mapminmax functions;
and 1.3, dividing the data set, and randomly dividing the training set and the test set according to the proportion of 8:2 to ensure the consistency of distribution.
Step 2, constructing a micro plastic abundance detection base model, which is specifically described as follows:
And 2.1, generating interaction items to obtain expression affecting the combined effect of the micro-plastic detection characteristics, wherein the micro-plastic detection characteristic matrix P_train (6 dimension) which is input as a model contains 6 characteristic values which are respectively characteristic 1-characteristic 6, and the interaction items generated according to the micro-plastic detection characteristic matrix are respectively (B6+B7), B6, (B5+B6), B7, (B5+B7) and (B5+B7)/B4. Where B4-B7 are each 400nm-2200nm band data from band remote sensing data from environmental monitoring. B4 is a wavelength band of 545nm to 560 nm, B5 is a wavelength band of 830 nm to 1250nm, B6 is a wavelength band of 1628 to 1652nm, and B7 is a wavelength band of 2105nm to 2135 nm. Then generating an interaction column of the characteristic 1 multiplied by the characteristic 2, and expanding the interaction column into 7 dimensions;
Step 2.2, performing interval variable transformation on the micro plastic detection features included in each generated interactive item, dynamically classifying the interval variable, firstly, automatically calculating the optimal bin boundary of a table by using histcount functions, selecting the first column of features of a training set, dividing the first column of features into five equal-width intervals, and then dividing the numerical value in the training set into corresponding bins by using discretize functions, wherein for example, the boundary is [0,2,4,6,8,10], and then the value of 3.5 is distributed to a second bin, namely the intervals [2,4]. Finally, preserving the box division boundary. And calling the pages parameter when the test set is binned, so that future information is prevented from being introduced.
Step 3, introducing a meta-learning fusion mode to perform parameter adjustment optimization of the original sample data set for the micro-plastic detection and training of the micro-plastic abundance detection base model, wherein the method is specifically described as follows:
Step 3.1, training a micro plastic abundance detection base model through a random forest model (RF model) to obtain a corresponding micro plastic abundance prediction random forest model, constructing a random forest model by using TreeBagger functions, defining TreeBagger parameters, wherein the main parameters are that the random forest number is 500, the splitting criterion of each tree is set to be 'regression', namely, the Mean Square Error (MSE) between an abundance prediction value and a true value detected by the micro plastic is minimum, the feature importance score is set to be 'OOBPredictorImportance', the importance of the micro plastic detection input data feature is evaluated by selecting samples (about 37% of data) which are not selected by the random forest, and selecting the samples as the optimized micro plastic detection input data feature according to the importance;
Step 3.2, training a micro-plastic detection original sample dataset through a gradient lifting model (Gradient Boosting, GB model) to obtain a micro-plastic abundance prediction gradient lifting model, constructing the gradient lifting model by using a fitrensemble function, setting a main setting parameter to be 80 weak learners (decision trees), setting a training method to be 'LSBoost', training the learners by a least square criterion, setting a decision target to minimize the sum of squares of residual errors of an abundance predicted value and a true value detected by the micro-plastic, setting the circulation times to be 50, setting the learning rate to be 0.1, and realizing the enhancement convergence and generalization effects of micro-plastic detection input data;
Step 3.3, training the micro-plastic detection original sample data set by using an RF model and a GB model to generate abundance predicted values to form a meta-feature matrix, and randomly and uniformly dividing the trained micro-plastic detection input data set obtained in step 3.1 and step 3.2 into 5 subsets, generating abundance predicted values (RF_ oof and GB_ oof) by using the RF model and the GB model through 5-fold cross validation, and constructing a 5-dimensional meta-feature matrix [ RF_ oof, GB_ oof, RF_ oof ×GB_ oof, RF_ oof 2,GB_oof2 ].
And 4, constructing a meta learner, training a meta model by using the meta learner, inputting the trained meta model into the characteristic engineering model for micro plastic detection and prediction, and executing the following steps:
Step 4.1, performing meta-feature standardization processing, specifically, calling zscore functions to realize meta-feature standardization as a standardized meta-feature matrix;
step 4.2, training a meta learner model, specifically, calling fitrensemble functions to construct a linear lifting (LSBoost) integrated model, using the meta feature matrix constructed in the step 3.3 as a feature value to be trained, setting a method= 'LSBoost', training the meta learner by using a least square criterion, wherein the aim is that the sum of the abundance predicted value and the true value residual error square of micro-plastic detection is minimum, setting a decision tree as 90, setting the learning rate as 0.05, preventing overfitting, setting the splitting frequency of the tree as 15, controlling the complexity, and setting the leaf node sample number as 10;
And 4.3, inputting the trained meta-learner model into the characteristic engineering model to perform micro-plastic detection prediction, and specifically, calling predict functions to obtain a micro-plastic abundance prediction value.
The method also designs model evaluation and visualization:
step 5, evaluating and visualizing the characteristic engineering model;
Step 5.1, carrying out data inverse normalization, specifically, reversely restoring the predicted results of the normalized training set and the testing set into an original range by using mapminmax functions;
And 5.2, calculating R 2 and RMSE, MAE, MSE of abundance predicted values and true values of the micro-plastic detection to evaluate the characteristic engineering model, wherein R 2 (determining coefficient, R-Squared) is used for measuring the interpretation capability of the model to the variance of a target variable, and the closer the value is 1, the better the fitting effect of the model to data is shown. MSE (mean square error, mean Squared Error) is the average of the squared differences of the abundance predicted and true values, reflecting the absolute magnitude of the predicted error. RMSE (root mean square error, root Mean Squared Error) is the square root of MSE, making the error dimension consistent with the target variable. MAE (mean absolute error ) average of absolute differences of abundance predictions and true values;
and 5.3, drawing a predicted result scatter diagram of the training set and the test set.
FIG. 2 shows the technical route of the micro plastic prediction method based on the fusion of feature engineering and meta-learning, and the specific details of the method of the invention are further described below with reference to FIG. 2:
1. Data preprocessing and feature engineering
The method comprises the steps of clearing null values in a data set and preprocessing the data.
1.1 Data cleaning, namely eliminating samples containing missing values or abnormal values, and ensuring the data quality.
1.2 Feature interactions generating product terms between features (e.g., feature 1x feature 2) to enhance nonlinear expression capabilities.
1.3, The sub-box processing is carried out, namely, discretization sub-box is carried out on the continuous characteristics, and the sub-box boundary is saved to adapt to the test set.
2. Base model training and optimization
This step uses the data from the end of the first step preprocessing to train the model. The two models are respectively and independently trained, so that the accuracy of the result is ensured.
2.1 Random forest model the minimum number of leaf nodes (3/5/10) and the number of trees (50/200/500) are optimized by grid search to minimize the training set RMSE. And obtaining a prediction result of the random forest model after operation.
2.2 Gradient lifting the tree model, namely adopting a Bagging method, setting the learning cycle number as 80, and improving the stability of the model. And obtaining a prediction result of gradient lifting after operation.
3. Meta learner construction
And taking the prediction results of the random forest and gradient lifting model as new characteristics to be input into a meta learner, training a linear lifting (LSBoost) integrated model, and further fusing the advantages of the model. And generating meta-characteristics by cross-checking, namely acquiring a predicted result and derivative characteristics (product term and square term) of the base model in the training set by using 5-fold cross-checking. And finally training the meta model, and iteratively optimizing the prediction result.
3.1 Cross-validation RF and GB abundance predictions (RF_oof, GB_ oof) were generated by 5-fold cross-validation as meta-feature inputs.
And 3.2, normalizing the meta-feature by Z-score to eliminate dimension influence.
3.3 Extend meta-feature dimensions including raw abundance predictors, interaction term (RF_ oof ×GB_ oof), square term (RF_ oof 2、GB_oof2).
Training a 3.4-element model, namely adopting LSBoost algorithm and combining a decision tree base learner (the maximum division number is 15 and the minimum leaf node number is 10), and iteratively optimizing a prediction result through 90 learning periods.
4. Model evaluation
And drawing a scatter diagram of the training set and the testing set, and calculating indexes such as R 2, MSE, MAE and the like. And evaluating the accuracy of the model according to the model index.
In summary, the invention combines the random forest and the gradient lifting model to construct the meta learner, thereby enhancing the utilization of meta features. Complex and subtle correlations between models are fully exploited. The precision and the universality are greatly improved.
Based on expert activation prediction mechanism without training cost, the model constructed by using the existing data can accurately invert the abundance of the microplastic in the Bohai sea. Experiments show that the model has the same distribution characteristics as practical for the micro plastic abundance from day to year.
And (3) verifying and analyzing, namely applying the trained model to the remote sensing image, and carrying out averaging to obtain an inversion chart of the distribution of the micro plastic abundance in a month of a certain year. Consistent with the actual measurement result, the inversion result only has errors in local areas, but the overall accuracy is within an acceptable range, and the inversion method can be used for space-time characteristic analysis research, so that the applicability is strong.
It should be noted that, while the present invention has been shown and described with reference to the particular exemplary embodiments thereof, it will be understood by those skilled in the art that the present invention is not limited to the above embodiments and various changes to the present invention fall within the scope of the present application.
Claims (8)
1. A microplastic prediction method based on fusion of feature engineering and meta-learning is characterized by comprising the following steps:
s1, acquiring micro-plastic detection data, selecting and combining through characteristic engineering, and extracting micro-plastic detection characteristics;
S2, constructing a micro-plastic abundance detection base model, inputting the micro-plastic detection characteristics into the micro-plastic abundance detection base model, and calculating to obtain a micro-plastic abundance prediction result;
S3, training the micro plastic abundance detection base model through a random forest algorithm and a gradient lifting algorithm respectively and independently to obtain a corresponding micro plastic abundance prediction random forest model and a micro plastic abundance prediction gradient lifting model;
S4, constructing a meta learner model, fusing the micro plastic abundance prediction random forest model and the micro plastic abundance prediction gradient lifting model through the meta learner model to obtain new micro plastic abundance detection characteristics, and inverting through cross verification to obtain a prediction result of the micro plastic abundance detection base model to generate micro plastic detection meta characteristics;
And S5, constructing a micro-plastic abundance detection meta-model, and performing iterative training on the micro-plastic abundance detection meta-model by utilizing the micro-plastic detection meta-features to obtain an optimized prediction result.
2. The method for predicting the microplastic based on the fusion of the characteristic engineering and the meta learning according to claim 1 is characterized in that S1 further comprises generating interaction items for detecting the characteristics of the microplastic, performing interval variable transformation, wherein the interaction items generated according to the microplastic detection characteristic matrix are respectively (B6+B7), B6, (B5+B6), B7, (B5+B7) and (B5+B7)/B4, B4-B7 are 400nm-2200nm wave band data of wave band remote sensing data from environmental monitoring, B4 is a wave band of 545 nm-560 nm, B5 is a wave band of 1230 nm-1250 nm, B6 is a wave band of 1628-1652 nm, B7 is a wave band of 2105 nm-2135 nm, and generating interaction columns of characteristic 1X characteristic 2, and the interaction columns are expanded into 7 dimensions;
And dynamically dividing the interval type variable into boxes, calculating an optimal box dividing boundary, selecting the first column of characteristics of the training set, dividing the first column of characteristics into five equal-width intervals, dividing the numerical values in the training set into corresponding boxes, and storing the box dividing boundary.
3. The method for predicting the microplastic based on the fusion of feature engineering and meta-learning according to claim 1, wherein S3 further comprises the steps of constructing a random forest model, training a microplastic detection original sample data set model through the random forest model, evaluating the importance of the microplastic detection input data features, and selecting the microplastic detection input data features as optimized microplastic detection input data features according to the importance;
Constructing a gradient lifting model, training a micro-plastic detection original sample data set through the gradient lifting model, training a learner by using a least square criterion, and making a decision target to minimize the sum of squares of residual errors of a micro-plastic abundance detection predicted result and a real result.
4. The method for predicting the micro-plastic based on the fusion of feature engineering and meta-learning according to claim 1, wherein the step S4 of obtaining the prediction result and the derivative features of the micro-plastic abundance detection base model through cross-validation, the step of generating micro-plastic abundance detection meta-features further comprises the steps of randomly and uniformly dividing micro-plastic detection input data trained by a random forest model and a gradient lifting model into 5 subsets, and carrying out 5-fold cross-validation to construct a meta-feature matrix [ RF_ oof, GB_ oof, RF_ oof ×GB_ oof, and RF_ oof 2,GB_oof2 ].
5. The method for predicting the micro plastic based on the fusion of feature engineering and meta-learning according to claim 1, wherein S5 further comprises the steps of constructing a micro plastic abundance detection meta-model, standardizing meta-features, training the micro plastic abundance detection meta-model, and obtaining a micro plastic abundance predicted value aiming at minimum sum of squares of abundance predicted results and real results of micro plastic detection.
6. The micro-plastic prediction method based on feature engineering and meta-learning fusion according to claim 1, wherein S3 further comprises the steps of constructing a meta-learner, generating a micro-plastic abundance detection random forest model and a micro-plastic abundance detection gradient lifting model through 5-fold cross validation, performing meta-feature standardization by taking the micro-plastic abundance detection random forest model and the micro-plastic abundance detection gradient lifting model as meta-features, expanding meta-feature dimensions comprising original predicted values, interaction terms RF_ oof ×GB_ oof, square terms RF_ oof2 and GB_ oof2, performing feature engineering meta-model training by adopting a LSBoost algorithm, and iteratively optimizing predicted results by combining a decision tree base learner.
7. The method for predicting microplastic based on fusion of feature engineering and meta-learning according to claim 1, further comprising visualization of a scatter diagram of predicted results of the training set and the test set.
8. The method for predicting the micro-plastic based on the fusion of feature engineering and meta-learning according to claim 1, wherein the method further comprises performing meta-learner model evaluation, namely reversely reducing the predicted results of the normalized training set and the testing set to an original range, and calculating a determination coefficient R 2 of an abundance predicted value and a real value of the micro-plastic detection, square root RMSE of a mean square error, mean absolute error MAE and mean square error MSE for evaluating the micro-plastic abundance detection meta-model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511011100.9A CN120853757A (en) | 2025-07-22 | 2025-07-22 | Microplastic prediction method based on fusion of feature engineering and meta-learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202511011100.9A CN120853757A (en) | 2025-07-22 | 2025-07-22 | Microplastic prediction method based on fusion of feature engineering and meta-learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN120853757A true CN120853757A (en) | 2025-10-28 |
Family
ID=97421108
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202511011100.9A Pending CN120853757A (en) | 2025-07-22 | 2025-07-22 | Microplastic prediction method based on fusion of feature engineering and meta-learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120853757A (en) |
-
2025
- 2025-07-22 CN CN202511011100.9A patent/CN120853757A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Avolio et al. | A comprehensive approach to analyzing community dynamics using rank abundance curves | |
| Hair Jr et al. | Evaluation of formative measurement models | |
| KR101811270B1 (en) | Method and system for checking goods | |
| Wadoux et al. | Perspectives on data‐driven soil research | |
| CN116597939A (en) | Medicine quality control management analysis system and method based on big data | |
| US20220374401A1 (en) | Determining domain and matching algorithms for data systems | |
| Ujjainia et al. | A crop recommendation system to improve crop productivity using ensemble technique | |
| Alamsyah et al. | Xgboost hyperparameter optimization using randomizedsearchcv for accurate forest fire drought condition prediction | |
| Chen et al. | Dynamic comprehensive quality assessment of post-harvest grape in different transportation chains using SAHP–CatBoost machine learning | |
| Rajbahadur et al. | Pitfalls analyzer: quality control for model-driven data science pipelines | |
| CN116881798A (en) | Conditional gracile causal analysis method based on variable selection and reverse time lag feature selection for complex systems such as weather | |
| CN120823047A (en) | An AI-based intelligent feasibility analysis and decision support system for investment projects | |
| CN120853757A (en) | Microplastic prediction method based on fusion of feature engineering and meta-learning | |
| CN119295120A (en) | A business opportunity recommendation method, device, equipment, medium and product for Class B enterprises | |
| Patil | Time Series Analysis and Stock Price Forecasting using Machine Learning Techniques | |
| Ding et al. | Metric information mining with metric attention to boost software defect prediction performance | |
| Mallamo et al. | Daily Engine Performance Trending Using Common Flight Regime Identification | |
| CN118706949B (en) | Disease detection method, device and computer equipment | |
| Yu et al. | Research on the design of a data mining-based financial audit model for financial multi-type data processing and audit trail discovery | |
| CN119808794B (en) | A big data intelligent analysis method and system based on AI | |
| Su et al. | Deep-learning-driven simulations of boundary layer clouds over the Southern Great Plains | |
| Carpenter | Machine Learning Brings Vast Core-Analysis Legacy Data to Life | |
| Utomo et al. | Enhancing Housing Price Prediction Accuracy Using Decision Tree Regression with Multivariate Real Estate Attributes | |
| Verezubova et al. | Eco-assessment of meat raw materials: A convolutional neural network approach to sustainable quality control | |
| CN120509957A (en) | Intelligent auxiliary credit giving method for consumption finance |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |