CN117436029A - Method and system for serial collaborative fusion of multiple large models - Google Patents
Method and system for serial collaborative fusion of multiple large models Download PDFInfo
- Publication number
- CN117436029A CN117436029A CN202311443875.4A CN202311443875A CN117436029A CN 117436029 A CN117436029 A CN 117436029A CN 202311443875 A CN202311443875 A CN 202311443875A CN 117436029 A CN117436029 A CN 117436029A
- Authority
- CN
- China
- Prior art keywords
- model
- models
- data set
- training
- fusion
- 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
- 230000004927 fusion Effects 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000007500 overflow downdraw method Methods 0.000 claims abstract 2
- 238000012549 training Methods 0.000 claims description 88
- 238000012360 testing method Methods 0.000 claims description 38
- 238000012545 processing Methods 0.000 claims description 37
- 238000011156 evaluation Methods 0.000 claims description 20
- 238000004140 cleaning Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000012800 visualization Methods 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 230000000007 visual effect Effects 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000013480 data collection Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/259—Fusion by voting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/27—Regression, e.g. linear or logistic regression
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a serial collaboration fusion method for a plurality of large models, which adopts a weighted average method to realize serial collaboration fusion of the large models, and comprises the steps S1-S6. Even if a certain model performs poorly in a particular situation, other models can still provide reliable prediction results; the complexity of the model can be gradually reduced, the risk of overfitting is reduced, and the generalization capability of the model can be improved; the multi-model serial collaboration fusion can provide richer model combination and feature selection modes, and the interpretability of the model is improved.
Description
Technical Field
The invention relates to the technical field of multi-model serial collaboration fusion, in particular to a method and a system for serial collaboration fusion of a plurality of large models.
Background
The multi-model serial collaboration fusion technology is used for modeling and predicting complex problems, based on the thought of serial collaboration, a plurality of large models are sequentially collaborated and fused according to a certain sequence, and a final prediction result is obtained through continuous iteration and interactive updating, so that the accuracy and generalization capability of the models are improved.
In the prior art, multiple models are required to be trained and optimized simultaneously through multi-model fusion, the complexity of training and optimizing is increased, multiple models are required to be operated simultaneously through multi-model fusion, more computing resources and memory are consumed, in multi-model fusion, dependency relationship exists among the models, and the output of each model can influence the input of the subsequent model. This may lead to accumulation of transfer errors between models, which in turn affects overall predictive performance, and therefore we need to propose a method and system for serial collaborative fusion of multiple large models.
Disclosure of Invention
The invention aims to provide a method and a system for serial collaborative fusion of a plurality of large models, wherein the multi-model serial collaborative fusion can fully exert the advantages of each model by connecting different types of models in series, make up the defect of a single model, reduce the influence of accidental errors of individual models on results and improve the robustness of the whole model through common efforts of the plurality of models. Even if a certain model performs poorly in a particular situation, other models can still provide reliable prediction results; the complexity of the model can be gradually reduced, the risk of overfitting is reduced, and the generalization capability of the model can be improved; the multi-model serial collaboration fusion can provide richer model combination and feature selection modes, and the interpretability of the model is increased, so that the problems in the background technology are solved.
In order to achieve the above purpose, the present invention provides the following technical solutions: a method for realizing serial collaboration fusion of a plurality of large models by adopting a weighted average method comprises the following steps:
s1, preparing a training data set test data set;
s2, selecting a plurality of basic models, and distributing weights for each basic model;
s3, training each basic model by using a training data set respectively to obtain a prediction result of the basic model;
s4, carrying out weighted average on the prediction result of the basic model according to the weight to obtain a final prediction result;
s5, predicting the test data set by using a final prediction result;
and S6, evaluating the fused model by using evaluation indexes of accuracy, recall and F1 values, and adjusting and optimizing the model according to the evaluation result.
Preferably, in step S1, preparing the training data set test data set includes the steps of:
a1, data collection: first, enough data needs to be collected for training and testing;
a2, data cleaning and pretreatment: cleaning and preprocessing the collected data, including removing repeated data, processing missing values and processing abnormal values;
a3, data division: dividing the cleaned and preprocessed data set into a training data set and a test data set;
a4, data set balance: the data set has the problem of unbalanced category, the sample number of partial category is small, and the problem of unbalanced data set can be processed by undersampling and oversampling methods.
Preferably, in step S2, in selecting the base model, the following factors are considered:
model type: selecting different types of models, and modeling data from different angles;
model performance: evaluating according to the accuracy of the model on a training data set or other evaluation indexes, and selecting a model with good performance;
model variability: the selected models have differences and are used for providing diversified prediction results;
model complexity: models of different complexity are selected, and a trade-off is made between the diversity of model predictions and the computational complexity of the model.
Preferably, in step S2, weights may be assigned to each base model in any of a uniform assignment, a performance-based assignment, a confidence-based assignment, and an experience-based assignment;
wherein, evenly distributing weights: adopting an arithmetic average mode to distribute equal weight to each basic model;
weight is assigned based on performance: distributing weights to the models according to the performance of each basic model on the training data set;
weights are assigned based on confidence: according to the confidence level or reliability of each basic model, a weight is distributed to the models;
weights are assigned based on experience: each base model is assigned a weight based on experience or domain knowledge.
Preferably, in step S3, the following detailed steps are included:
b1, selecting a proper basic model according to the properties of tasks and data;
b2, training a basic model: for each basic model, training by using a training data set, training the model by using a corresponding training method and an optimization algorithm according to the type and algorithm of the model, inputting the input characteristics and the target variable of the training data set into the model in the training process, and updating the parameters of the model according to the difference between the output of the model and the target variable;
b3, obtaining a prediction result of the basic model: predicting a training data set by using the trained basic models, inputting the input features of the training data set into each basic model, and obtaining a prediction result of the model;
and B4, repeating the step B2 and the step B3 until all the basic models are trained and the prediction results are obtained.
Preferably, in step S4, the prediction result of each basic model is multiplied by the weight corresponding to the prediction result, and all the weighted results are added to obtain a final prediction result;
the final prediction is equal to Σ (base model prediction) where Σ represents summing the predictions of all base models.
Preferably, in step S5, the following detailed steps are included:
c1, performing feature scaling, standardization and coding feature processing on a test data set;
and C2, inputting the test data set subjected to the feature processing into a final model to obtain a prediction result.
The invention also provides a serial collaboration fusion system for the large models, which comprises a main control module, a data processing module, a model training module, a model prediction module, a model fusion module, a performance evaluation module and a visualization and result display module, wherein the data processing module, the model training module, the model prediction module, the model fusion module, the performance evaluation module and the visualization and result display module are all connected with the main control module.
Preferably, the data processing module: the method comprises the steps of preprocessing original data and characteristic processing, wherein the preprocessing comprises data cleaning, repeated data removal, missing value processing and abnormal value processing, and the characteristic processing comprises characteristic selection, characteristic transformation, characteristic coding and data standardization;
the model training module: for training a plurality of independent basis models, including linear models, tree models, different types of models of neural networks;
the model prediction module: the method comprises the steps of predicting test data, predicting the test data by using a trained basic model, and generating a prediction result of the basic model;
the model fusion module: the prediction results of the plurality of basic models are fused;
the performance evaluation module: the method is used for evaluating the performance of the fusion model, calculating the indexes of the accuracy rate, recall rate and F1 value of the model, and helping a user to select an optimal model fusion strategy;
the visualization and result display module: the method is used for visually displaying the processes and results of data processing, model training and model fusion, providing visual charts and visual tools, and helping users understand and analyze the performance and results of the models.
Compared with the prior art, the invention has the beneficial effects that:
the multi-model serial collaborative fusion can fully exert the advantages of each model by connecting different types of models in series, make up the defect of a single model, reduce the influence of accidental errors of individual models on results through the common effort of a plurality of models, and improve the robustness of the whole model. Even if a certain model performs poorly in a particular situation, other models can still provide reliable prediction results; the complexity of the model can be gradually reduced, the risk of overfitting is reduced, and the generalization capability of the model can be improved; the multi-model serial collaboration fusion can provide richer model combination and feature selection modes, and the interpretability of the model is improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of the present invention for preparing a training data set test data set;
FIG. 3 is a block flow diagram of the model training of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1-3, the present invention provides a technical solution: a method for realizing serial collaboration fusion of a plurality of large models by adopting a weighted average method comprises the following steps:
s1, preparing a training data set test data set;
s2, selecting a plurality of basic models, and distributing weights for each basic model;
s3, training each basic model by using a training data set respectively to obtain a prediction result of the basic model;
s4, carrying out weighted average on the prediction result of the basic model according to the weight to obtain a final prediction result;
s5, predicting the test data set by using a final prediction result;
and S6, evaluating the fused model by using evaluation indexes of accuracy, recall and F1 values, and adjusting and optimizing the model according to the evaluation result.
The stacking method can also be used for realizing serial collaborative fusion of a plurality of large models, and comprises the following steps:
d1, preparing a training data set and a test data set;
d2, selecting a plurality of basic models, and dividing the training data set into a plurality of subsets;
d3, training each subset by using different basic models to obtain a prediction result of the basic model,
D4, taking a prediction result of the basic model as a new characteristic, and constructing a meta model;
and D5, predicting the test data set by using the meta model.
The voting method can also be used for realizing serial collaborative fusion of a plurality of large models, and comprises the following steps:
e1, preparing a training data set and a test data set;
e2, selecting a plurality of basic models;
e3, training each basic model by using a training data set to obtain a prediction result of the basic model;
e4, voting the predicted result of the basic model, and selecting the category with the largest occurrence number as the final predicted result;
and E5, predicting the test data set by using the final prediction result.
The learning method can also be used for realizing serial collaborative fusion of a plurality of large models, and the method comprises the following steps:
z1, preparing a training data set and a test data set;
z2, selecting a plurality of basic models, and dividing the training data set into a plurality of subsets;
z3, training each subset by using different basic models respectively to obtain a prediction result of the basic model;
z4, learning and fusing the prediction result of the basic model by using a meta model to obtain a final prediction result;
and Z5, predicting the test data set by using the final prediction result.
In step S1, preparing a training data set test data set includes the steps of:
a1, data collection: first, enough data needs to be collected for training and testing; the data may be obtained from a public data set or collected by itself according to the characteristics of the task.
A2, data cleaning and pretreatment: cleaning and preprocessing the collected data, including removing repeated data, processing missing values and processing abnormal values; pretreatment steps such as feature selection, feature encoding, data standardization and the like can also be carried out.
A3, data division: dividing the cleaned and preprocessed data set into a training data set and a test data set; common partitioning methods include random partitioning, proportional partitioning, cross-validation, and the like. Typically, a large portion of the dataset is used for training and a small portion is used for testing.
Training data set: the training data set is a data set for training a model. It contains the input features and corresponding target variables (tags). The training data set should be representative and capable of covering various situations of the task.
Test dataset: the test dataset is a dataset for evaluating the performance of the model. It also contains input features and corresponding target variables (labels), but the target variables are typically not used by the model for training. The test data set should be independent of the training data set, and reflect the performance of the model in practical applications.
A4, data set balance: the data set has the problem of unbalanced category, the sample number of partial category is small, and the problem of unbalanced data set can be processed by undersampling and oversampling methods.
In step S2, in selecting the base model, the following factors are considered:
model type: selecting different types of models, such as a linear model, a tree model, a neural network and the like, and modeling data from different angles;
model performance: evaluating according to the accuracy of the model on a training data set or other evaluation indexes, and selecting a model with good performance;
model variability: the selected models have differences and are used for providing diversified prediction results;
model complexity: models of different complexity are selected, and a trade-off is made between the diversity of model predictions and the computational complexity of the model.
In step S2, weights may be assigned to each base model in any one of a uniform assignment, a performance-based assignment, a confidence-based assignment, and an experience-based assignment;
wherein, evenly distributing weights: adopting an arithmetic average mode to distribute equal weight to each basic model;
weight is assigned based on performance: distributing weights to the models according to the performance of each basic model on the training data set; performance may be measured in terms of accuracy, F1 score, etc.
Weights are assigned based on confidence: according to the confidence level or reliability of each basic model, a weight is distributed to the models; confidence may be calculated based on the probability of model predictions or other confidence measures.
Weights are assigned based on experience: each base model is assigned a weight based on experience or domain knowledge. This may require some preliminary evaluation and comparison of the model.
In step S3, the following detailed steps are included:
b1, selecting a proper basic model according to the properties of tasks and data; different types of models may be selected, such as linear models, tree models, neural networks, etc.
B2, training a basic model: for each basic model, training by using a training data set, training the model by using a corresponding training method and an optimization algorithm according to the type and algorithm of the model, inputting the input characteristics and the target variable of the training data set into the model in the training process, and updating the parameters of the model according to the difference between the output of the model and the target variable;
b3, obtaining a prediction result of the basic model: predicting a training data set by using the trained basic models, inputting the input features of the training data set into each basic model, and obtaining a prediction result of the model; the prediction result may be a classification prediction result of a classification model or a numerical prediction result of a regression model.
And B4, repeating the step B2 and the step B3 until all the basic models are trained and the prediction results are obtained.
Through the steps, the prediction result of each basic model on the training data set can be obtained. These predictions will be used for subsequent model fusion or performance evaluation.
In step S4, multiplying the prediction result of each basic model by the weight corresponding to the prediction result, and adding all weighted results to obtain a final prediction result;
the final prediction is equal to Σ (base model prediction) where Σ represents summing the predictions of all base models.
And carrying out post-processing on the final prediction result according to task requirements. For example, the classification problem may be probability normalized, thresholded for classification, etc.
Through the steps, the prediction result of the basic model can be weighted and averaged according to the weight, and the final prediction result is obtained. The weighted average method can integrate the prediction capability of a plurality of basic models, and improve the performance and the robustness of the whole model.
In step S5, the following detailed steps are included:
c1, performing feature scaling, standardization and coding feature processing on a test data set; in particular, the training data set also performs feature scaling, normalization, and encoding feature processing.
And C2, inputting the test data set subjected to the feature processing into a final model to obtain a prediction result. For classification problems, classification can be performed according to the final prediction result; for regression problems, the final prediction result may be directly used as a prediction value.
Through serial fusion of multiple models, the advantages of different models can be fully utilized, and the performance and the robustness of the models are improved.
The invention also provides a serial collaboration fusion system for the large models, which comprises a main control module, a data processing module, a model training module, a model prediction module, a model fusion module, a performance evaluation module and a visualization and result display module, wherein the data processing module, the model training module, the model prediction module, the model fusion module, the performance evaluation module and the visualization and result display module are all connected with the main control module.
And a data processing module: the method comprises the steps of preprocessing original data and characteristic processing, wherein the preprocessing comprises data cleaning, repeated data removal, missing value processing and abnormal value processing, and the characteristic processing comprises characteristic selection, characteristic transformation, characteristic coding and data standardization;
model training module: for training a plurality of independent basis models, including linear models, tree models, different types of models of neural networks; the module generally comprises functions of model selection, super-parameter tuning, model training and the like.
Model prediction module: the method comprises the steps of predicting test data, predicting the test data by using a trained basic model, and generating a prediction result of the basic model;
model fusion module: the prediction results of the plurality of basic models are fused; the fusion may be performed using voting, averaging, weighted averaging, stacking, etc. The module can provide different fusion strategies and parameter configurations to meet different requirements.
Performance evaluation module: the method is used for evaluating the performance of the fusion model, calculating the indexes of the accuracy rate, recall rate and F1 value of the model, and helping a user to select an optimal model fusion strategy;
and a visualization and result display module: the method is used for visually displaying the processes and results of data processing, model training and model fusion, providing visual charts and visual tools, and helping users understand and analyze the performance and results of the models.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (9)
1. A method for serial collaborative fusion of a plurality of large models is characterized by adopting a weighted average method to realize serial collaborative fusion of a plurality of large models, and comprises the following steps:
s1, preparing a training data set test data set;
s2, selecting a plurality of basic models, and distributing weights for each basic model;
s3, training each basic model by using a training data set respectively to obtain a prediction result of the basic model;
s4, carrying out weighted average on the prediction result of the basic model according to the weight to obtain a final prediction result;
s5, predicting the test data set by using a final prediction result;
and S6, evaluating the fused model by using evaluation indexes of accuracy, recall and F1 values, and adjusting and optimizing the model according to the evaluation result.
2. The method for serial collaborative fusion of multiple large models according to claim 1, wherein: in step S1, preparing a training data set test data set includes the steps of:
a1, data collection: first, enough data needs to be collected for training and testing;
a2, data cleaning and pretreatment: cleaning and preprocessing the collected data, including removing repeated data, processing missing values and processing abnormal values;
a3, data division: dividing the cleaned and preprocessed data set into a training data set and a test data set;
a4, data set balance: the data set has the problem of unbalanced category, the sample number of partial category is small, and the problem of unbalanced data set can be processed by undersampling and oversampling methods.
3. The method for serial collaborative fusion of multiple large models according to claim 1, wherein: in step S2, in selecting the base model, the following factors are considered:
model type: selecting different types of models, and modeling data from different angles;
model performance: evaluating according to the accuracy of the model on a training data set or other evaluation indexes, and selecting a model with good performance;
model variability: the selected models have differences and are used for providing diversified prediction results;
model complexity: models of different complexity are selected, and a trade-off is made between the diversity of model predictions and the computational complexity of the model.
4. The method for serial collaborative fusion of multiple large models according to claim 1, wherein: in step S2, weights may be assigned to each base model in any one of a uniform assignment, a performance-based assignment, a confidence-based assignment, and an experience-based assignment;
wherein, evenly distributing weights: adopting an arithmetic average mode to distribute equal weight to each basic model;
weight is assigned based on performance: distributing weights to the models according to the performance of each basic model on the training data set;
weights are assigned based on confidence: according to the confidence level or reliability of each basic model, a weight is distributed to the models;
weights are assigned based on experience: each base model is assigned a weight based on experience or domain knowledge.
5. The method for serial collaborative fusion of multiple large models according to claim 1, wherein: in step S3, the following detailed steps are included:
b1, selecting a proper basic model according to the properties of tasks and data;
b2, training a basic model: for each basic model, training by using a training data set, training the model by using a corresponding training method and an optimization algorithm according to the type and algorithm of the model, inputting the input characteristics and the target variable of the training data set into the model in the training process, and updating the parameters of the model according to the difference between the output of the model and the target variable;
b3, obtaining a prediction result of the basic model: predicting a training data set by using the trained basic models, inputting the input features of the training data set into each basic model, and obtaining a prediction result of the model;
and B4, repeating the step B2 and the step B3 until all the basic models are trained and the prediction results are obtained.
6. The method for serial collaborative fusion of multiple large models according to claim 1, wherein: in step S4, multiplying the prediction result of each basic model by the weight corresponding to the prediction result, and adding all weighted results to obtain a final prediction result;
the final prediction is equal to Σ (base model prediction) where Σ represents summing the predictions of all base models.
7. The method for serial collaborative fusion of multiple large models according to claim 1, wherein: in step S5, the following detailed steps are included:
c1, performing feature scaling, standardization and coding feature processing on a test data set;
and C2, inputting the test data set subjected to the feature processing into a final model to obtain a prediction result.
8. A serial collaboration fusion system of multiple large models, based on a serial collaboration fusion method of multiple large models according to any one of claims 1-7, characterized in that: the system comprises a main control module, a data processing module, a model training module, a model prediction module, a model fusion module, a performance evaluation module and a visualization and result display module, wherein the data processing module, the model training module, the model prediction module, the model fusion module, the performance evaluation module and the visualization and result display module are all connected with the main control module.
9. The multiple large model serial collaboration fusion system of claim 8, wherein: the data processing module: the method comprises the steps of preprocessing original data and characteristic processing, wherein the preprocessing comprises data cleaning, repeated data removal, missing value processing and abnormal value processing, and the characteristic processing comprises characteristic selection, characteristic transformation, characteristic coding and data standardization;
the model training module: for training a plurality of independent basis models, including linear models, tree models, different types of models of neural networks;
the model prediction module: the method comprises the steps of predicting test data, predicting the test data by using a trained basic model, and generating a prediction result of the basic model;
the model fusion module: the prediction results of the plurality of basic models are fused;
the performance evaluation module: the method is used for evaluating the performance of the fusion model, calculating the indexes of the accuracy rate, recall rate and F1 value of the model, and helping a user to select an optimal model fusion strategy;
the visualization and result display module: the method is used for visually displaying the processes and results of data processing, model training and model fusion, providing visual charts and visual tools, and helping users understand and analyze the performance and results of the models.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311443875.4A CN117436029A (en) | 2023-11-01 | 2023-11-01 | Method and system for serial collaborative fusion of multiple large models |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311443875.4A CN117436029A (en) | 2023-11-01 | 2023-11-01 | Method and system for serial collaborative fusion of multiple large models |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117436029A true CN117436029A (en) | 2024-01-23 |
Family
ID=89545845
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311443875.4A Pending CN117436029A (en) | 2023-11-01 | 2023-11-01 | Method and system for serial collaborative fusion of multiple large models |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117436029A (en) |
-
2023
- 2023-11-01 CN CN202311443875.4A patent/CN117436029A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107992976B (en) | Hot topic early development trend prediction system and prediction method | |
CN112364975A (en) | Terminal operation state prediction method and system based on graph neural network | |
CN111553543A (en) | Power load prediction method based on TPA-Seq2Seq and related assembly | |
CN112819523B (en) | Marketing prediction method combining inner/outer product feature interaction and Bayesian neural network | |
CN112967088A (en) | Marketing activity prediction model structure and prediction method based on knowledge distillation | |
CN112288137A (en) | LSTM short-term load prediction method and device considering electricity price and Attention mechanism | |
CN111027672A (en) | Time sequence prediction method based on interactive multi-scale recurrent neural network | |
CN113255844A (en) | Recommendation method and system based on graph convolution neural network interaction | |
Guo et al. | Dynamic Fuzzy Logic Control of Genetic Algorithm Probabilities. | |
CN115114128A (en) | Satellite health state evaluation system and evaluation method | |
CN116340726A (en) | Energy economy big data cleaning method, system, equipment and storage medium | |
CN113657678A (en) | Power grid power data prediction method based on information freshness | |
CN113033898A (en) | Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network | |
CN115201394B (en) | Multi-component transformer oil chromatography online monitoring method and related device | |
CN111539558A (en) | Power load prediction method adopting optimized extreme learning machine | |
CN117436029A (en) | Method and system for serial collaborative fusion of multiple large models | |
CN115619028A (en) | Clustering algorithm fusion-based power load accurate prediction method | |
CN115169521A (en) | Graph neural network interpretation method for keeping prediction sequence and structure dependency relationship | |
CN113111308B (en) | Symbolic regression method and system based on data-driven genetic programming algorithm | |
US20230297837A1 (en) | Method for automated determination of a model compression technique for compression of an artificial intelligence-based model | |
CN114726751A (en) | Intelligent early warning method, system, equipment and storage medium for resource quality monitoring | |
CN113837481A (en) | Financial big data management system based on block chain | |
CN113723692A (en) | Data processing method, apparatus, device, medium, and program product | |
Hsieh | Employing data mining technique to achieve the parameter optimization based on manufacturing intelligence | |
Bigaud et al. | Decision-making through a fuzzy hybrid AI system for selection of a third-party operations and maintenance provider |
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 |