Pollutant prediction method and device based on evolutionary learning strategy, equipment and storage medium

Abstract

The invention discloses a pollutant prediction method, device and equipment based on an evolutionary learning strategy, and a storage medium, and relates to the technical field of pollutant prediction, and the method comprises the steps: carrying out the data exception processing of pollutant data, obtaining cleaned data, carrying out the feature conversion and time series data noise reduction of the cleaned data, obtaining feature information, and carrying out the data exception processing of the feature information; and identifying the feature information through a target pollutant prediction model to obtain a pollutant prediction result. Through multi-model fusion optimization, parameters are dynamically adjusted by using an evolutionary learning algorithm, a target pollutant prediction model is constructed, and the prediction precision of pollutants is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pollutant prediction, and in particular to a pollutant prediction method, device, equipment and storage medium based on an evolutionary learning strategy. Background Art

[0002] Traditional pollutant prediction methods rely primarily on statistical models and simple machine learning algorithms. These methods perform well when processing relatively simple and linear data. However, as data volumes increase and pollutant sources become more complex, traditional methods are gradually showing their limitations. Statistical models struggle to effectively process different types of data, especially when the data comes from a wide range of sources and in diverse forms. Furthermore, they are built on linear assumptions and have limited ability to handle complex nonlinear relationships. Furthermore, when processing high-dimensional data, they are prone to overfitting, resulting in reduced generalization capabilities. Therefore, a method that uses multiple models working together and adaptive optimization mechanisms to predict pollutant data is needed to improve the accuracy of pollutant predictions. Summary of the Invention

[0003] The main purpose of this application is to provide a pollutant prediction method, device, equipment and storage medium based on evolutionary learning strategy, aiming to solve the technical problem of how to improve the accuracy of pollutant prediction.

[0004] To achieve the above objectives, this application proposes a pollutant prediction method based on an evolutionary learning strategy, the method comprising: Obtain pollutant data; Performing data anomaly processing on the pollutant data to obtain cleaned data; Performing feature conversion and time series data noise reduction on the cleaned data to obtain feature information; Identify the characteristic information through a target pollutant prediction model to obtain pollutant prediction results, including pollutant data prediction results, pollutant category prediction results, and pollutant nonlinear and high-dimensional characteristic data prediction results; Before the step of identifying the characteristic information through a target pollutant prediction model to obtain a pollutant prediction result, the method includes: Optimizing the preset hybrid expert model according to the multi-level knowledge fusion strategy to obtain a preset pollutant prediction model, wherein the preset hybrid expert model includes a linear regression expert model, a support vector regression expert model, a random forest expert model, and a deep neural network expert model; The preset pollutant prediction model is optimized according to the evolutionary learning algorithm strategy to obtain a target pollutant prediction model.

[0005] In one embodiment, the step of optimizing the preset hybrid expert model according to the multi-level knowledge fusion strategy to obtain the preset pollutant prediction model includes: Obtain task requirements for pollutant prediction; According to the task requirements, select the corresponding optimization fusion method in the multi-level knowledge fusion strategy, which includes weighted average fusion, weighted voting fusion and weighted projection fusion; Performing a fusion process on the preset hybrid expert model according to the optimization fusion method to obtain an optimized preset hybrid expert model; The optimized preset hybrid expert model is used as a preset pollutant prediction model.

[0006] In one embodiment, the step of selecting a corresponding optimized fusion method in the multi-level knowledge fusion strategy according to the task requirements includes: When the task requirement is pollutant data prediction, the optimal fusion method is to give priority to weighted average fusion, and the specific expression formula is: in, is the predicted value of pollutants after fusion, For the Pollutant prediction values ​​of expert models, For the The weight of the expert model, and satisfy , is the total number of expert models; When the task requirement is pollutant category prediction, the optimal fusion method is to prioritize weighted voting fusion, and the specific expression formula is: in, is the pollutant prediction category, For the Category of class, is the total number of pollutant categories, For the Expert model prediction category The probability of For the The weight of the expert model; When the task requires the prediction of nonlinear and high-dimensional feature data, the optimal fusion method is to give priority to weighted projection fusion. The specific expression formula is: in, is the predicted value of pollutants after projection, is a high-dimensional vector, represents the projection function, is the optimized output function.

[0007] In one embodiment, the step of optimizing the preset pollutant prediction model according to the evolutionary learning algorithm strategy to obtain the target pollutant prediction model includes: Obtaining hyperparameters and model prediction accuracy corresponding to the expert model of the preset hybrid expert model; Encoding the hyperparameters and weight parameters in the multi-level knowledge fusion strategy to obtain individual gene values; Based on the prediction accuracy of the model, the fitness function is constructed to obtain the fitness value. The fitness function is in the form of: in, is the actual value of fitness, is the predicted value of fitness, is the number of data; Optimizing the fitness value and the individual gene value according to an evolutionary operation strategy and a multi-objective optimization strategy to obtain optimized hyperparameters and optimized fusion weights; The preset pollutant prediction model is updated according to the optimized hyperparameters and the optimized fusion weights to obtain a target pollutant prediction model.

[0008] In one embodiment, the step of optimizing the fitness value and the individual gene value according to the evolutionary operation strategy and the multi-objective optimization strategy to obtain optimized hyperparameters and optimized fusion weights includes: By sorting the fitness values ​​and the individual gene values, selecting individuals corresponding to preset fitness values ​​and preset individual gene values; Performing crossover and mutation operations on the individuals to obtain updated individuals; Optimizing the fitness value of the update individual through a multi-objective optimization strategy to obtain a Pareto optimal solution set; A preset solution is selected from the Pareto optimal solution and used as the optimized hyperparameter and the optimized fusion weight.

[0009] In one embodiment, the step of optimizing the fitness values ​​of the update individuals by a multi-objective optimization strategy to obtain a Pareto optimal solution set includes: sorting the fitness values ​​of the updated individuals to obtain a plurality of crowding distance levels; According to the screening of the crowding distance level, the fitness values ​​of multiple updated individuals corresponding to the preset level are selected as Pareto optimal solutions; Based on combining multiple Pareto optimal solutions, a Pareto optimal solution set is obtained.

[0010] In one embodiment, after the step of identifying the characteristic information using a target pollutant prediction model to obtain a pollutant prediction result, the method further includes: Real-time monitoring and acquisition of updated pollutant data; Inputting the updated pollutant data into the target pollutant prediction model to obtain an evaluation performance; When the evaluation performance does not meet the preset prediction requirements, the hyperparameters and fusion weights of the target pollutant prediction model are adjusted until the evaluation performance meets the preset prediction requirements.

[0011] In addition, to achieve the above objectives, the present application also proposes a pollutant prediction device based on an evolutionary learning strategy, the pollutant prediction device based on an evolutionary learning strategy comprising: Acquisition module, used to obtain pollutant data; A data cleaning module is used to obtain cleaned data by filling missing values ​​and detecting and correcting outliers on the pollutant data; A data conversion module is used to perform feature conversion and time series data noise reduction on the cleaned data to obtain feature information; A result module is used to identify the characteristic information through the target pollutant prediction model to obtain pollutant prediction results, which include pollutant data prediction results, pollutant category prediction results, and pollutant nonlinear and high-dimensional characteristic data prediction results; A construction module is used to optimize the preset hybrid expert model according to a multi-level knowledge fusion strategy to obtain a preset pollutant prediction model. The preset hybrid expert model includes a linear regression expert model, a support vector regression expert model, a random forest expert model and a deep neural network expert model; the preset pollutant prediction model is optimized according to an evolutionary learning algorithm strategy to obtain a target pollutant prediction model.

[0012] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable medium and stores a computer program. When the computer program is executed by a processor, it implements the steps of the pollutant prediction method based on the evolutionary learning strategy as described above.

[0013] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the steps of the pollutant prediction method based on the evolutionary learning strategy as described above.

[0014] This application obtains pollutant data, processes it for data anomalies, obtains cleaned data, performs feature conversion and time series noise reduction on the cleaned data, obtains characteristic information, and uses this characteristic information to identify the target pollutant prediction model to obtain pollutant prediction results. Through multi-model fusion optimization, an evolutionary learning algorithm is used to dynamically adjust parameters to construct a target pollutant prediction model, thereby improving pollutant prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0016] Figure 1 This is a flow chart of the first embodiment of the pollutant prediction method based on the evolutionary learning strategy of this application; Figure 2 This is a flow chart of the second embodiment of the pollutant prediction method based on the evolutionary learning strategy of this application; Figure 3 This is a flow chart of the third embodiment of the pollutant prediction method based on the evolutionary learning strategy of this application; Figure 4 This is a schematic diagram of the module structure of the pollutant prediction device based on the evolutionary learning strategy according to an embodiment of the present application; Figure 5 Schematic diagram of the device structure of the hardware operating environment involved in the pollutant prediction method based on evolutionary learning strategy in the embodiment of the present application.

[0017] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0018] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0019] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0020] Accurately predicting the concentrations and trends of these pollutants is crucial for developing effective environmental management strategies, pollution control measures, and public health protection initiatives. However, traditional pollutant prediction methods face drawbacks when faced with data diversity, nonlinear relationships, and high-dimensional features. These issues primarily manifest themselves in the following aspects: First, pollutant data often contains missing values ​​and outliers, which directly impact the accuracy and reliability of subsequent model training. For example, in air quality monitoring, data at certain points in time may be lost or erroneous readings may occur due to equipment failure or external interference. Second, the relationship between pollutant concentrations and their influencing factors is often complex and nonlinear, posing a challenge to traditional prediction models based on linear assumptions. Furthermore, with advances in monitoring technology, the dimensionality of environmental data continues to increase, posing a significant challenge in extracting effective information from it. Furthermore, a single machine learning model is insufficient to address all types of pollutant prediction tasks. Each model has its own strengths and limitations. For example, linear regression excels at handling simple linear relationships, while deep neural networks are better suited to capturing complex nonlinear patterns. Therefore, effectively combining the strengths of multiple models to build a prediction system that can adapt to diverse pollutant characteristics and handle complex data structures has become a key research direction.

[0021] Therefore, the main solution of the embodiment of the present application is: to obtain cleaned data by performing data anomaly processing on pollutant data, to perform feature conversion and time series data denoising on the cleaned data to obtain feature information, to identify the feature information through the target pollutant prediction model, and to obtain pollutant prediction results.

[0022] Based on this, the embodiment of the present application provides a pollutant prediction method based on evolutionary learning strategy, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the pollutant prediction method based on the evolutionary learning strategy of this application.

[0023] In this embodiment, the pollutant prediction method based on the evolutionary learning strategy includes steps S10 to S40: Step S10: Acquire pollutant data.

[0024] It should be noted that in modern society, with the improvement of environmental protection awareness and the development of technology, the sources of pollutant data are becoming increasingly diverse and complex. These data usually come from various monitoring stations, satellite remote sensing and mobile sensors, covering pollutant concentration information in various environmental media such as atmosphere, water, soil, etc. For example, in air quality monitoring, pollutant data include PM2.5, PM10, sulfur dioxide ( ), nitrogen oxides ( ) and other key indicators; for water quality monitoring, pollutant data focuses on chemical oxygen demand (COD), ammonia nitrogen ( ), heavy metal content, etc.

[0025] Step S20 , performing data anomaly processing on the pollutant data to obtain cleaned data.

[0026] It's important to note that missing values ​​are a common anomaly in environmental monitoring data, often caused by sensor failures, network transmission issues, or human error. To address this issue, employing an appropriate imputation method is crucial. For example, KNN imputation uses the average of the nearest neighboring samples to the target sample to fill missing values; regression imputation predicts missing values ​​by building a regression model based on other complete features; and mean imputation directly uses the average of all known values ​​of the feature. In particular, for time series data, interpolation using data from previous and subsequent time points can be considered to maintain temporal continuity.

[0027] Furthermore, outliers are detected and corrected. Outliers are caused by measurement errors or extreme events. If not handled, they will have a negative impact on model training. Commonly used methods include statistical analysis methods such as interquartile range (IQR) and Z-score. The IQR method identifies outliers by calculating the gap between the first and third quartiles. Any value that is 1.5 times lower than the lower quartile or 1.5 times higher than the upper quartile is considered an outlier. The Z-score rule calculates the standard score for each data point. If the absolute value of the Z-score of a point exceeds a certain threshold, the point is considered an outlier. Once the outliers are identified, different strategies can be used to correct them, such as replacing them with the mean of nearby points, deleting abnormal records, or making specific adjustments based on domain knowledge. After completing the above steps, a cleaned dataset is obtained.

[0028] Step S30: Perform feature conversion and time series data noise reduction on the cleaned data to obtain feature information.

[0029] It should be noted that feature transformation involves converting raw data into a form more suitable for processing by machine learning models. This step includes a variety of techniques. For example, by applying logarithmic transformations or Box-Cox transformations, datasets with skewed distributions can be effectively processed, making the data more consistent with a normal distribution, thereby improving the learning efficiency of the model. Furthermore, appropriate transformation methods can be selected based on specific application scenarios. For example, when dealing with concentration-dependent pollutants, since their concentration changes tend to grow or decay exponentially, appropriate nonlinear transformations can better capture this relationship.

[0030] Secondly, denoising time series data is a crucial step in optimizing data characteristics along the temporal dimension. Environmental pollutant monitoring typically generates large amounts of time series data, which are susceptible to short-term fluctuations or random noise, which can obscure true trends and cyclical changes. To address this issue, signal processing techniques such as wavelet transforms or Kalman filtering can be employed. Wavelet transforms analyze signals at different scales, helping to isolate useful information and remove high-frequency noise. Kalman filtering, based on state estimation theory, uses a recursive algorithm to smooth noise in the data, making it particularly suitable for time series prediction of dynamic systems. After these feature transformations and denoising processes, we obtain feature information that not only preserves the core characteristics of the original data but also enhances the model's understanding of complex nonlinear relationships. This lays the foundation for subsequent model training and optimization processes to more accurately reflect pollutant behavior in the real environment, thereby achieving more precise prediction results.

[0031] Step S40: identifying the characteristic information through a target pollutant prediction model to obtain a pollutant prediction result.

[0032] It should be noted that in this embodiment, the above-mentioned prediction results include predictions for pollutant data, pollutant categories, and nonlinear, high-dimensional feature data. Once the feature information is prepared, it is fed into a hybrid expert model that has been previously optimized using an evolutionary learning algorithm. This model is composed of multiple expert models, each of which independently predicts the data based on its specific strengths. For example, a linear regression expert model may handle simple linear relationships, while a deep neural network expert model focuses on capturing complex nonlinear patterns. To maximize the strengths of each model and minimize their limitations, a multi-level knowledge fusion strategy, such as weighted average fusion, weighted voting fusion, and weighted projection fusion, is employed to integrate the outputs of the expert models. Each fusion strategy is flexibly applied based on specific task requirements. For numerical prediction tasks, weighted average fusion is preferred; for classification problems, weighted voting fusion is used; and for data with complex, nonlinear, high-dimensional features, weighted projection fusion is used. This integrated approach not only improves prediction accuracy but also enhances the robustness and adaptability of the system. Ultimately, based on these optimized models and fusion strategies, predictions for pollutant concentration or type under specific environmental conditions can be output. This not only helps to understand the current environmental pollution situation in a timely manner, but also provides a scientific basis for formulating effective pollution control measures.

[0033] Specifically, before step S40, it also includes: optimizing based on a preset hybrid expert model and a multi-level knowledge fusion strategy to obtain a preset pollutant prediction model. It should be noted that this process first relies on building a hybrid expert model that includes multiple machine learning models, and each model makes independent predictions for a specific type of data feature or task. The above-mentioned preset hybrid expert models include a linear regression expert model, a support vector regression expert model, a random forest expert model, and a deep neural network expert model. Specifically, the linear regression expert model (LR) is suitable for pollutant prediction scenarios with simple, linear relationships. By standardizing the input data, a least squares regression model is constructed, assuming that there is a linear relationship between pollutant concentrations and environmental factors. The formula is: in, is the intercept, is the regression coefficient of each feature, The current number The eigenvalues ​​of Indicates the total number of data. The basic idea of ​​the LR model is to assume that the prediction target With the input eigenvalue The LR model is used to predict the relationship between pollutant concentrations and environmental factors. It is characterized by strong interpretability and low computational complexity.

[0034] The support vector regression expert model (SVR) is good at capturing pollutant data with complex nonlinear relationships. SVR maps low-dimensional data to high-dimensional space through kernel functions, thereby effectively capturing nonlinear patterns. The specific model formula is: in, is the kernel function, is the Lagrange multiplier, is the bias term, Indicates the data, The above formula is derived from the regression form (SVR) of the support vector machine (SVM), and is obtained by the kernel function The input data is mapped from the original space to a high-dimensional feature space so that a linear regressor can be constructed in the space to capture the nonlinear relationship between input and output. The SVR expert model is used to process complex nonlinear patterns in pollutant data and can effectively characterize nonlinear and non-stationary pollution time series changes. The SVR expert model described by the formula can embed the pollutant prediction problem into a high-dimensional kernel space to achieve modeling of the relationship between nonlinear features and pollutant concentrations. Kernel function , RBF kernel, polynomial kernel to adapt to the changing patterns of different pollutant data, Sparsity means that only a portion of the training samples (support vectors) will affect the prediction, which greatly reduces the computational cost when modeling large-scale environmental data. The derivation process for the above formula is: In the pollutant prediction task, through the known historical data , through the regression function, the regression function is made to be Can predict the value as close to the true value as possible In the SVR expert model, the regression function can be expressed as: in, Indicates that the feature vector Mapping function to high-dimensional feature space, is the transpose of the weight vector in high-dimensional space, Represents the bias term. Then, by introducing the Lagrange multiplier and the weight vector of the high-dimensional space , the above regression function can be expressed as: in, is the Lagrange multiplier, represents the inner product, Indicates the data, Represents the total number of data. Then replace the inner product with the kernel function and record , substitute The final function representation can be obtained.

[0035] The Random Forest Expert Model (RF) reduces overfitting by integrating multiple decision trees. The specific formula is: in, For the A tree, For the Tree pairs sample This formula describes the prediction mechanism of the Random Forest (RF) model, which is conceived as A decision tree trained on a random subset of features To reduce variance and improve generalization ability. The prediction function of each tree is The final output of the forest is the average of all tree predictions. Based on the bagging principle, ideally, the expected mean squared error of the forest is equal to the mean of the individual tree errors minus the covariance of inter-tree predictions, indicating that diversity improves accuracy. The RF expert model utilizes a multi-tree averaging mechanism to reduce sensitivity to noise, making it suitable for pollutant prediction scenarios with a large number of feature dimensions and sample sizes.

[0036] The deep neural network expert model (DNN) can automatically extract deep features from high-dimensional data and learn the deep nonlinear relationship between data through multi-layer perceptrons. The specific model formula is: in, is the weight matrix of each layer, is the bias matrix, is the activation function. During the training and evaluation phase of the expert model, in order to ensure that each model not only performs well on its training data, but also maintains good generalization capabilities on unseen data, it is crucial to use cross-validation techniques such as K-fold cross-validation. K-fold cross-validation randomly divides the entire dataset into K subsets or "folds", and then performs K iterative training and testing processes, each time using one of the subsets as the test set and the remaining K-1 subsets as the training set. This method helps to evaluate the stability and robustness of the model and ensure its performance consistency on different data subsets. When evaluating the performance of each expert model, a variety of performance indicators are used to comprehensively measure the effectiveness of the model. Mean squared error (MSE) is a commonly used evaluation criterion for regression problems. It calculates the average of the sum of the squares of the differences between the predicted values ​​and the actual values. The lower the MSE, the higher the prediction accuracy of the model. Coefficient of determination ( ) is used to measure the ability of the model to explain variability, and its value ranges from 0 to 1. The closer it is to 1, the better the model fit. In addition, other indicators such as mean absolute error (MAE) can be considered to obtain a more comprehensive understanding of model performance. In addition, based on the hybrid expert model system, a multi-level knowledge fusion strategy is used to integrate the prediction results of each expert model to maximize the advantages of different models and overcome their limitations. This strategy includes but is not limited to weighted average fusion, weighted voting fusion, and weighted projection fusion. Weighted average fusion assigns weights to each expert model and then performs a weighted average of their prediction results, which is particularly suitable for numerical prediction tasks. Weighted voting fusion is mainly used for classification problems. It assigns weights according to the classification ability of each expert model and determines the final category through weighted voting. For complex nonlinear regression tasks, weighted projection fusion maps the outputs of multiple expert models to a low-dimensional space and improves prediction performance by optimizing the fusion weights and mapping functions.

[0037] Furthermore, the preset pollutant prediction model is optimized according to an evolutionary learning algorithm strategy to obtain a target pollutant prediction model. It should be noted that in this embodiment, the evolutionary learning algorithm is a genetic algorithm (GA). GA is a search heuristic algorithm that simulates natural selection and genetic mechanisms and is used to solve optimization and search problems. This provides a relatively simple approach to solving complex optimization problems. The process can be divided into the following main stages: population initialization, individual encoding, fitness evaluation, selection, crossover, mutation, new population formation, and termination condition determination. First, a set of solutions is randomly generated as the initial population, each of which is called an individual. These individuals represent a potential solution to the problem. During the individual encoding stage, the hyperparameters of each expert model (e.g., the kernel function type for SVM, the number of trees for RF, the number of layers for DNN) and the fusion weights between the expert models are encoded to form the individual's genome. These encodings can use binary encoding, real number encoding, or rule-based encoding schemes to facilitate subsequent genetic operations. Next, a fitness function is defined, which evaluates the performance of the individual based on the accuracy of pollutant predictions (such as mean squared error (MSE) or mean absolute error (MAE)). A higher fitness indicates that the model structure and parameter settings corresponding to that individual are closer to the optimal solution. During the evolutionary process, new populations of individuals are continuously generated through selection, crossover, and mutation. Selection selects outstanding individuals based on their fitness values. Common methods include roulette wheel selection and tournament selection. Crossover generates new individuals using single-point or two-point crossover to explore new combinations in the hyperparameter space. Mutation introduces small perturbations to increase population diversity and prevent premature convergence to local optima. Furthermore, in multi-objective optimization scenarios, algorithms such as NSGA-II can be used to balance model complexity and prediction accuracy, ensuring that the model is both accurate and not overfitting. When the maximum number of iterations is reached, a satisfactory solution is found, or the population's fitness no longer improves, the resulting best individual is output as the approximate optimal solution to the problem. If this condition is not met, the algorithm returns to the next iteration. After multiple iterations, the evolutionary learning algorithm will find an optimized set of hyperparameter configurations and fusion strategies, resulting in a more accurate pollutant prediction model, which is then used as the target pollutant prediction model. This model can not only more accurately capture the complex patterns and nonlinear relationships in the data, but also dynamically adapt to environmental changes and maintain long-term stability and efficiency. Finally, the performance of the optimized model is evaluated through verification on the test set, and the indicators before and after the improvement, such as MSE and determination coefficient, are compared. and MAE, etc., to confirm the optimization effect.

[0038] Furthermore, after step S40, a dynamic adaptive and online learning mechanism is introduced. Specifically, updated pollutant data is monitored and acquired in real time. This updated pollutant data is then fed into the target pollutant prediction model to obtain an evaluation performance. If the evaluation performance does not meet the preset prediction requirements, the hyperparameters and fusion weights of the target pollutant prediction model are adjusted until the evaluation performance meets the preset prediction requirements. This mechanism enables the system to receive new pollutant data in real time and dynamically adjust model parameters based on this data, ensuring that it remains efficient and accurate under changing environmental conditions. Specifically, the online learning mechanism allows the model to continuously update its parameters during operation without requiring the entire model to be retrained. For example, if newly monitored data indicates abnormal fluctuations in the concentration of a particular pollutant, the system can immediately incorporate this information and adjust the prediction model accordingly to more accurately reflect current environmental conditions. Furthermore, by using evolutionary algorithms (such as genetic algorithms (GA) or particle swarm optimization (PSO)), the weights and fusion strategies of each expert model can be dynamically optimized based on the latest data, further enhancing the system's adaptive capabilities. This dynamic adaptive and online learning capability is crucial for addressing complex environmental pollution issues. It not only improves the accuracy and reliability of the prediction model, but also enables the system to quickly respond to sudden pollution incidents and provide early warning information in a timely manner.

[0039] This embodiment provides a pollutant prediction method based on an evolutionary learning strategy. This method performs data anomaly processing on pollutant data to obtain cleaned data. Feature conversion and time series noise reduction are then performed on the cleaned data to obtain characteristic information. This characteristic information is then used to identify the target pollutant prediction model, resulting in pollutant prediction results. Through multi-model fusion optimization, an evolutionary learning algorithm dynamically adjusts parameters to construct a target pollutant prediction model, improving pollutant prediction accuracy.

[0040] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 2 The pollutant prediction method based on the evolutionary learning strategy further includes steps S201 to S204 before step S40: Step S201: Obtaining task requirements for pollutant prediction.

[0041] It should be noted that the task requirements for obtaining pollutant prediction involve multiple aspects. In this embodiment, they are divided into three categories according to different application scenarios: obtaining pollutant data, obtaining pollutant categories, and predicting nonlinear and high-dimensional feature data.

[0042] Specifically, the demand for obtaining pollutant data is mainly focused on the accurate prediction of the trend of changes in pollutant concentrations in specific environmental media. For example, in the field of air quality monitoring, it may be necessary to predict the concentrations of pollutants such as PM2.5 and SO2 in the next few days. This requires the system to be able to process large amounts of real-time data and have high precision and rapid response capabilities. The demand for obtaining pollutant categories focuses on identifying the specific type of pollutant or determining whether it exceeds the standard. For example, in water quality monitoring, it is necessary to determine whether there is heavy metal pollution or excessive organic matter in the water body, and further analyze its source. For the prediction needs of nonlinear, high-dimensional feature data, such as the analysis of multi-pollutant interactions in complex atmospheric or water environments, it is usually necessary to process high-dimensional data sets with complex nonlinear relationships.

[0043] Step S202: Select the corresponding optimization fusion method in the multi-level knowledge fusion strategy according to the task requirements.

[0044] It should be noted that the above-mentioned optimization fusion methods include weighted average fusion, weighted voting fusion and weighted projection fusion.

[0045] Step S203 , performing fusion processing on the preset hybrid expert model according to the optimized fusion method to obtain an optimized preset hybrid expert model.

[0046] It should be noted that, when the task requirement is pollutant data prediction, the optimal fusion method is to give priority to weighted average fusion, and the specific expression formula is: in, is the predicted value of pollutants after fusion, For the Pollutant prediction values ​​of expert models, For the The weight of the expert model, and satisfy , is the total number of expert models. This formula reflects the idea of ​​weighted average fusion, that is, according to the prediction value of each expert model Credibility weight , and their linear combination gets the final prediction value The weight can be obtained by normalizing the inverse of the historical error of the model or optimizing it using an evolutionary algorithm. First, each expert model needs to be weighted. Specifically, in this embodiment, the weight allocation method adopts a static allocation based on the model error, which mainly relies on the prediction error on the historical data to determine the weight of each expert model. Specifically, models with lower errors will receive higher weights, which can be achieved by calculating indicators such as mean square error (MSE) or mean absolute error (MAE). The specific formula for static weight allocation is: in, For the The mean square error of the model, Indicates the The error of each model is determined by the weights. However, environmental conditions and data characteristics change over time, necessitating a method for dynamically adjusting weights. By monitoring the prediction error of each expert model in real time and dynamically adjusting its weight based on its latest performance, the system can better adapt to changing data patterns. A common technique is the sliding window method, which calculates the performance of each expert model over a recent period of time, thereby assigning higher weights to models with better recent performance. This method is particularly suitable for processing non-stationary data, that is, data distributions that change over time. After the weights are assigned, the predictions of each expert model are first collected. Then, corresponding weights are assigned based on the historical or real-time errors of these models. Finally, the final prediction result is calculated using a weighted average formula. This method not only improves prediction accuracy but also enhances the system's adaptability and robustness, ensuring reliable prediction results under various conditions.

[0047] Furthermore, when the task requirement is pollutant category prediction, the optimal fusion method is to prioritize weighted voting fusion, and the specific expression formula is: in, is the pollutant prediction category, For the Category of class, is the total number of pollutant categories, For the Expert model prediction category The probability of For the The weight of each expert model. The formula expresses the weighted voting classification fusion process. Each expert model outputs The predicted probability , assign weights to expert models based on classification accuracy or confidence , sum the weighted probabilities of each category, and select the category with the maximum probability as the final prediction result. The derivation idea comes from Bayesian decision theory: if the model predictions are independent, then maximizing the sum of weighted probabilities can approximately minimize the classification error rate. In this patented system, this method is used for pollutant category discrimination (such as pollution level identification), which can balance generalization ability and decision credibility. First, it is necessary to assign weights to each expert model. Specifically, in this embodiment, the weight assignment method adopts an assignment based on classification accuracy. The specific formula is: in, For the The classification accuracy of each expert model is often not optimal. However, when faced with rapidly changing data or non-stationary distributions, a dynamic weighting strategy based on prediction confidence is more effective. This approach adjusts weights in real time based on the expert model's prediction confidence (e.g., classification probability) for the current sample. When an expert model has a higher confidence level for a particular category, it is given a higher weight. For example, when dealing with multi-class pollutant identification, if an expert model predicts a significantly higher probability for a particular pollutant type than for other categories, its judgment on that category is considered more reliable. After weighting is assigned, the classification results or category probabilities output by each expert model are collected. Next, corresponding weights are calculated based on the expert model's classification accuracy or the prediction confidence for the current sample. Finally, these weights and category probabilities are combined to determine the final predicted category through weighted voting or other fusion mechanisms. This dynamic weighting approach not only improves the system's flexibility and adaptability but also more accurately captures subtle changes in the data, providing more reliable support for environmental monitoring and management.

[0048] Furthermore, when the task requires the prediction of nonlinear and high-dimensional feature data, the optimal fusion method is to give priority to weighted projection fusion. The specific expression formula is: in, is the predicted value of pollutants after projection, is a high-dimensional vector, represents the projection function, is the optimized output function, Indicates the The weights of the expert models. The formula is used for the fusion of nonlinear, high-dimensional prediction tasks. The high-dimensional output of each model Pre-projection function Map to a potential space (can be dimensionality reduction mapping or kernel space mapping), and then according to the weight Linear combination, and finally through the output function Producing predictions. The derivation and conception refer to the combined ideas of feature transformation and regression. The projection step before fusion avoids the information loss or noise amplification problems caused by directly weighting inconsistent high-dimensional outputs. This method is used to integrate the modeling advantages of different models for high-dimensional spatiotemporal features. Weighted projection fusion includes linear projection and nonlinear projection. Directly using linear weighted projection is a simple and effective strategy, particularly suitable for scenarios where the predicted values ​​are scalar. By assigning weights to each expert model and weightedly summing their predictions, the final prediction result is obtained. This method is computationally efficient and suitable for handling relatively straightforward, linear data relationships. However, when faced with data with complex nonlinear and high-dimensional features, nonlinear projection becomes more appropriate. A common approach is to use kernel methods (such as RBF kernels) to map the predictions of each expert model into a high-dimensional space, fuse them within this space, and then project them back to the original space. Deep autoencoders can also be used to learn the optimal fused representation of the predicted values, automatically extracting features to capture deep patterns in the data. This approach not only handles complex nonlinear relationships but also effectively mitigates the impact of the curse of dimensionality. After obtaining the weights, the projection function and output function are defined, the prediction results of each expert model are calculated and projected into the target space, and then the weights are optimized to calculate the final prediction value based on the projected results.

[0049] Through the above series of optimization and fusion processes, we can obtain an optimized preset hybrid expert model. This model not only inherits the advantages of each expert model, but also improves the predictive ability and adaptability of the overall system through a multi-level knowledge fusion strategy.

[0050] Step S204: Using the optimized preset hybrid expert model as a preset pollutant prediction model.

[0051] This embodiment obtains pollutant prediction task requirements, selects a corresponding optimization fusion method from a multi-level knowledge fusion strategy based on the task requirements, fuses the preset hybrid expert model according to the optimized fusion method, and obtains an optimized preset hybrid expert model. This optimized preset hybrid expert model is used as the preset pollutant prediction model. By selecting an appropriate multi-level knowledge fusion strategy based on the task requirements and optimizing the preset hybrid expert model, a high-precision pollutant prediction model is obtained, improving the accuracy and adaptability of pollutant prediction.

[0052] Based on the first embodiment of the present application, in the third embodiment of the present application, the same or similar contents as those in the first embodiment can be referred to the above introduction and will not be described in detail later. Figure 3 The pollutant prediction method based on the evolutionary learning strategy further includes steps S301 to S305 before step S40: Step S301: Obtain hyperparameters and model prediction accuracy corresponding to the expert model of the preset hybrid expert model.

[0053] It should be noted that before making model predictions, the hyperparameters of each expert model must be clarified. For example, in the linear regression (LR) model, no specific hyperparameters need to be set, but in the support vector regression (SVR) model, the kernel function type, regularization parameter, and kernel coefficient need to be determined; for random forest (RF), the number of trees and the maximum depth need to be set; and in deep neural networks (DNN), the number of layers, the number of neurons in each layer, the activation function type, and the learning rate are all important hyperparameters. These hyperparameters directly affect the performance of the model, so choosing appropriate values ​​is crucial. The optimal hyperparameter combination can be explored through grid search, random search, or more advanced Bayesian optimization methods. In order to evaluate the performance of each expert model, cross-validation techniques such as K-fold cross-validation are usually used. In this process, the data is divided into K subsets, the model is trained on K-1 subsets and tested on the remaining subset, and this process is repeated K times, using a different subset as the test set each time. In this way, a series of performance indicators can be calculated, including mean squared error (MSE), mean absolute error (MAE), and coefficient of determination ( ), etc., are used to quantify the prediction accuracy of the model.

[0054] Step S302 , encoding the hyperparameters and weight parameters in the multi-level knowledge fusion strategy to obtain individual gene values.

[0055] It should be noted that for hyperparameters, real number encoding can be used to represent continuous variables (such as learning rate, regularization coefficient, etc.), while integer encoding can be used for discrete variables (such as the number of trees, number of layers, etc.). For example, in support vector regression (SVR), the kernel function type can be represented by integer encoding (1 represents a linear kernel, 2 represents an RBF kernel, etc.), while parameters such as regularization parameters and kernel coefficients are represented by real number encoding. Similarly, the number of trees and maximum depth in a random forest can also be represented by corresponding encoding methods. Next, for the weight parameters in the multi-level knowledge fusion strategy, real number encoding can be used to ensure that the sum of all weights is equal to 1. For example, in weighted average fusion, the weight of each expert model can be represented by a real number vector, satisfying , is the total number of expert models. In this way, the entire individual genome consists of the encoding of these hyperparameters and weight parameters, forming a complete solution, based on which the corresponding individual gene value can be obtained.

[0056] This encoding method enables evolutionary algorithms to search for optimal configurations in a vast solution space. The individual genomes not only contain the optimal hyperparameter settings for each expert model, but also reflect the optimal collaboration between different models, thereby improving the predictive accuracy and adaptability of the hybrid expert model.

[0057] Step S303: Evaluate based on the model prediction accuracy, construct a fitness function, and obtain a fitness value.

[0058] It should be noted that the design of the fitness function needs to be closely centered around specific task requirements and performance indicators. The fitness function is constructed based on the actual value and the predicted value for evolutionary algorithm optimization. The fitness function is in the form of: in, is the actual value of fitness, is the predicted value of fitness, is the number of data points. This formula defines how individual fitness is calculated, using mean squared error (MSE) to measure the performance of an individual (expert model configuration and fusion weight combination) on the prediction task. This formula is derived from a common error metric for regression tasks and is suitable as an optimization objective for genetic algorithms or other evolutionary learning strategies because it is continuous and comparable, and minimizing MSE is equivalent to maximizing prediction accuracy. This fitness function drives the evolutionary learning process, continuously screening and optimizing model parameters and fusion strategies, enabling iterative improvement of pollutant prediction models under multi-objective constraints. For a single sample: in, Denotes the square of the prediction error, expressed as a positive number and penalizing large deviations. samples, and take the average: Therefore, directly use MSE as Fitness (the smaller the better).

[0059] Step S304 , optimizing the fitness value and the individual gene value according to the evolutionary operation strategy and the multi-objective optimization strategy to obtain optimized hyperparameters and optimized fusion weights.

[0060] It should be noted that the overall performance of the model is improved by dynamically adjusting hyperparameters and fusion weights through the selection, crossover, and mutation of evolutionary operation strategies and the introduction of multi-objective optimization strategies.

[0061] Furthermore, step S304 also includes sorting fitness values ​​and individual gene values ​​to select individuals corresponding to a preset fitness and preset individual gene value. Specifically, the selection operation selects individuals with outstanding performance based on fitness values ​​to participate in the next generation's reproduction process. Common selection methods include roulette wheel selection and tournament selection. For example, in roulette wheel selection, individuals with higher fitness and higher individual gene values ​​are more likely to be selected, ensuring the inheritance of superior traits. The individuals are then processed through crossover and mutation operations to obtain updated individuals. Specifically, the crossover operation combines gene fragments from two parent individuals to generate new offspring. Single-point or two-point crossover methods can effectively explore new combinations in the hyperparameter space. For example, combining the regularization parameters of one parent individual in a support vector regression (SVR) model with the kernel coefficient parameters of another parent individual can produce a more optimal configuration. The mutation operation then increases population diversity by randomly perturbing individual genes, avoiding regression into local optima. For example, making small changes to the layer encoding of a deep neural network (DNN) for a particular individual can generate a new, superior population. The fitness values ​​of the update individuals are then optimized using a multi-objective optimization strategy to obtain a Pareto-optimal solution set. Preset solutions are selected from the Pareto-optimal solution set and used as optimized hyperparameters and fusion weights. Specifically, to balance model complexity and prediction accuracy, a multi-objective optimization strategy is employed. In this embodiment, the NSGA-II method is used. The fitness values ​​of the update individuals are sorted to obtain multiple crowding distance hierarchies. Based on the crowding distance hierarchies, the fitness values ​​of multiple update individuals corresponding to the preset hierarchies are selected as the Pareto-optimal solutions. These multiple Pareto-optimal solutions are then combined to obtain a Pareto-optimal solution set. Specifically, the fitness values ​​are sorted, and all update individuals are sorted according to their performance on different objectives. This means that individuals are divided into different hierarchies, with the first hierarchical level containing all non-dominated solutions (i.e., the Pareto front), the second hierarchical level containing individuals dominated only by individuals in the first hierarchical level, and so on. Next, the crowding distance is calculated, which measures the density of individuals within their hierarchical level. Specifically, for each individual, the sum of its distances to its neighbors along each objective dimension is calculated. A higher crowding distance indicates that the individual is sparse in its neighborhood and therefore more likely to be retained, maintaining population diversity. Then, based on the crowding distance level, multiple updated individuals from a preset level (usually the first few levels) are selected as candidate Pareto-optimal solutions. These individuals not only perform well in their respective non-dominated layers but also exhibit good diversity due to their larger crowding distances. Multiple Pareto-optimal solutions are combined to form a comprehensive Pareto-optimal solution set.This set contains solutions that achieve the best balance between different objectives, and then finds the Pareto optimal solution between multiple objectives (such as MSE and model complexity), so that the final optimized hyperparameters and fusion weights can ensure high accuracy without causing overfitting.

[0062] Through the above series of evolutionary operations and multi-objective optimization strategies, the hyperparameters and fusion weights are continuously iteratively optimized, and finally the optimized preset hybrid expert model is obtained.

[0063] Step S305 : updating the preset pollutant prediction model according to the optimized hyperparameters and the optimized fusion weights to obtain a target pollutant prediction model.

[0064] It should be noted that by applying the optimal solution obtained by the evolutionary algorithm and multi-objective optimization strategy to various expert models and fusion strategies, a target pollutant prediction model with better performance is constructed.

[0065] Specifically, each expert model is updated using the optimized optimal hyperparameters. For example, in the SVR model, the kernel function type, regularization parameter, and kernel coefficient are adjusted; in the RF model, the number of trees and the maximum depth are updated; for the DNN, the number of layers, the number of neurons per layer, the activation function type, and the learning rate are adjusted. These optimized hyperparameters can improve the performance of each expert model in specific tasks. Next, the multi-level knowledge fusion strategy is reconfigured using the optimized fusion weights. If the task requires numerical prediction, the weighted average fusion method is used to ensure that models with lower errors have higher weights; for classification tasks, weighted voting fusion is used to make models with high classification accuracy play a greater role; and when dealing with complex nonlinear relationships, weighted projection fusion is applied, and the fusion results are further optimized through deep autoencoders or kernel methods. Through the above steps, an optimized pollutant prediction model is obtained.

[0066] Furthermore, after obtaining the optimized pollutant prediction model, the process also includes obtaining pollutant validation data. This pollutant validation data should be independent of the test dataset to ensure fair and objective evaluation. Validation data typically comes from actual monitoring stations, satellite remote sensing, or other reliable environmental monitoring methods, covering a wide range of pollutant types and their corresponding environmental conditions. The target pollutant prediction model is then tested based on the validation data to obtain test results. The model parameters are then iteratively updated based on the optimized hyperparameters and fusion weights until a maximum number of iterations is reached or the test results meet the preset detection requirements. The target pollutant prediction model is then obtained. Specifically, the validation data is input into the model, and prediction results are generated. The model's performance is comprehensively evaluated by comparing the predicted values ​​with the true values, resulting in test results. Based on the test results, if the model fails to meet the preset detection requirements, the iterative update phase begins. In this phase, the model parameters are dynamically adjusted using an evolutionary algorithm or particle swarm optimization, combined with the optimized hyperparameters and fusion weights. After each iteration, the fitness function is recalculated, and the best-performing individuals are selected as the basis for the next generation. The iterative update process continues until the maximum number of iterations is reached or the model's performance on the validation set meets the preset detection criteria. Each iteration aims to further improve the model's predictive accuracy and robustness, ensuring its efficiency and stability in complex real-world environments. When the model ultimately meets all requirements, a prediction model for the target pollutant is obtained. This model not only boasts enhanced predictive accuracy and stability but also addresses a variety of complex real-world application scenarios, providing strong technical support for environmental protection.

[0067] This embodiment obtains the hyperparameters and prediction accuracy of the preset hybrid expert model, uses evolutionary operations and multi-objective optimization strategies to optimize the hyperparameters and fusion weights, and finally updates the preset model to obtain the target pollutant prediction model. Pollutants are predicted using the optimized target pollutant prediction model, thereby improving the accuracy of pollutant prediction and enhancing adaptability.

[0068] This application also provides a pollutant prediction device based on evolutionary learning strategy, please refer to Figure 4 , the device comprises: The acquisition module 10 is used to acquire pollutant data.

[0069] The data cleaning module 20 is used to obtain cleaned data by performing data anomaly processing on the pollutant data.

[0070] The data conversion module 30 is used to perform feature conversion and time series data noise reduction on the cleaned data to obtain feature information.

[0071] The result module 40 is used to identify the characteristic information through the target pollutant prediction model to obtain pollutant prediction results. The above prediction results include pollutant data prediction results, pollutant category prediction results and pollutant nonlinear and high-dimensional characteristic data prediction results.

[0072] Construction module 50 is used to optimize the preset hybrid expert model according to the multi-level knowledge fusion strategy to obtain a preset pollutant prediction model. The above-mentioned preset hybrid expert model includes a linear regression expert model, a support vector regression expert model, a random forest expert model and a deep neural network expert model; the preset pollutant prediction model is optimized according to the evolutionary learning algorithm strategy to obtain a target pollutant prediction model.

[0073] The pollutant prediction device based on an evolutionary learning strategy provided in this application, which employs the pollutant prediction method based on an evolutionary learning strategy described in the aforementioned embodiments, can address the technical problem of improving the accuracy of pollutant predictions. Compared to the prior art, the beneficial effects of the pollutant prediction device based on an evolutionary learning strategy provided in this application are the same as those of the pollutant prediction method based on an evolutionary learning strategy described in the aforementioned embodiments. Other technical features of the pollutant prediction device based on an evolutionary learning strategy are the same as those disclosed in the aforementioned embodiments and are not further elaborated here.

[0074] In one embodiment, the construction module 50 is also used to obtain task requirements for pollutant prediction; according to the task requirements, the corresponding optimization fusion method in the multi-level knowledge fusion strategy is selected, and the optimization fusion method includes weighted average fusion, weighted voting fusion and weighted projection fusion; the preset hybrid expert model is fused according to the optimization fusion method to obtain an optimized preset hybrid expert model; and the optimized preset hybrid expert model is used as the preset pollutant prediction model.

[0075] In one embodiment, the construction module 50 is also used to optimize the fusion method by giving priority to weighted average fusion when the task requirement is pollutant data prediction; optimize the fusion method by giving priority to weighted voting fusion when the task requirement is pollutant category prediction; and optimize the fusion method by giving priority to weighted projection fusion when the task requirement is nonlinear, high-dimensional feature data prediction.

[0076] In one embodiment, the construction module 50 is also used to obtain the hyperparameters and model prediction accuracy corresponding to the expert model of the preset hybrid expert model; encode the hyperparameters and weight parameters in the multi-level knowledge fusion strategy to obtain individual gene values; evaluate based on the model prediction accuracy, construct a fitness function, and obtain a fitness value; optimize the fitness value and individual gene value according to the evolutionary operation strategy and the multi-objective optimization strategy to obtain optimized hyperparameters and optimized fusion weights; update the preset pollutant prediction model according to the optimized hyperparameters and the optimized fusion weights to obtain a target pollutant prediction model.

[0077] In one embodiment, the construction module 50 is further used to select individuals corresponding to preset fitness values ​​and preset individual gene values ​​by sorting fitness values ​​and individual gene values; perform crossover and mutation operations on the individuals to obtain updated individuals; optimize the fitness values ​​of the updated individuals through a multi-objective optimization strategy to obtain a Pareto optimal solution set; and select a preset solution from the Pareto optimal solution and use it as the optimized hyperparameter and the optimized fusion weight.

[0078] In one embodiment, the construction module 50 is further used to sort the fitness values ​​of the updated individuals to obtain multiple crowding distance levels; based on the screening of the crowding distance levels, the fitness values ​​of multiple updated individuals corresponding to the preset levels are selected as Pareto optimal solutions; based on the combination of multiple Pareto optimal solutions, a Pareto optimal solution set is obtained.

[0079] In one embodiment, the result module 40 is also used to monitor and obtain updated pollutant data in real time; input the updated pollutant data into the target pollutant prediction model to obtain evaluation performance; when the evaluation performance does not meet the preset prediction requirements, adjust the hyperparameters and fusion weights of the target pollutant prediction model until the evaluation performance meets the preset prediction requirements.

[0080] The present application provides a pollutant prediction device based on an evolutionary learning strategy, and the pollutant prediction device based on the evolutionary learning strategy includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the pollutant prediction method based on the evolutionary learning strategy in the above-mentioned embodiment one.

[0081] Reference below Figure 5, which shows a schematic diagram of the structure of a pollutant prediction device based on an evolutionary learning strategy suitable for implementing embodiments of the present application. The pollutant prediction device based on an evolutionary learning strategy in embodiments of the present application can include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The pollutant prediction device based on the evolutionary learning strategy shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0082] like Figure 5 The pollutant prediction device based on an evolutionary learning strategy may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in ROM 1002 (Read Only Memory) or programs loaded from storage device 1003 into RAM 1004 (Random Access Memory). RAM 1004 also stores various programs and data required for the operation of the pollutant prediction device based on an evolutionary learning strategy. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to the I / O interface 1006: input devices 1007 including, for example, a touch screen, touchpad, keyboard, mouse, image sensor, microphone, accelerometer, gyroscope, etc.; output devices 1008 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1003 including, for example, a magnetic tape, hard disk, etc.; and communication devices 1009. Communication devices 1009 can allow the pollutant prediction device based on the evolutionary learning strategy to communicate wirelessly or wired with other devices to exchange data. Although the figure shows a pollutant prediction device based on the evolutionary learning strategy with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems may be implemented or have alternatively.

[0083] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes program code for executing the method described in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are performed.

[0084] The pollutant prediction device based on an evolutionary learning strategy provided in this application, which employs the pollutant prediction method based on an evolutionary learning strategy described in the aforementioned embodiment, can address the technical problem of improving the accuracy of pollutant predictions. Compared to the prior art, the beneficial effects of the pollutant prediction device based on an evolutionary learning strategy provided in this application are the same as those of the pollutant prediction method based on an evolutionary learning strategy described in the aforementioned embodiment. Other technical features of this pollutant prediction device based on an evolutionary learning strategy are the same as those disclosed in the aforementioned embodiment and are not further elaborated here.

[0085] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0086] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0087] The present application provides a computer-readable medium having computer-readable program instructions (i.e., computer program) stored thereon for performing calculations to obtain computer-readable program instructions for executing the pollutant prediction method based on the evolutionary learning strategy in the above-mentioned embodiment.

[0088] The computer-readable medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. Calculations are performed in this embodiment to obtain a machine-readable medium that may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0089] The computer-readable medium may be included in the pollutant prediction device based on the evolutionary learning strategy; or may exist independently without being assembled into the pollutant prediction device based on the evolutionary learning strategy.

[0090] The computer-readable medium carries one or more programs that, when executed by a pollutant prediction device based on an evolutionary learning strategy, enable the pollutant prediction device based on an evolutionary learning strategy to write computer program code for performing the operations of the present application in one or more programming languages, or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0091] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0092] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0093] The computer-readable medium provided in this application stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned pollutant prediction method based on an evolutionary learning strategy. This computer-readable medium addresses the technical problem of improving the accuracy of pollutant predictions. Compared to the prior art, the beneficial effects of the computer-readable medium provided in this application are similar to those of the pollutant prediction method based on an evolutionary learning strategy provided in the aforementioned embodiments, and are not further elaborated here.

[0094] The present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the above-mentioned pollutant prediction method based on evolutionary learning strategy.

[0095] The computer program product provided in this application can solve the technical problem of improving the accuracy of pollutant prediction. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the pollutant prediction method based on evolutionary learning strategy provided in the above embodiment, and will not be elaborated here.

[0096] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A pollutant prediction method based on evolutionary learning strategy, characterized in that: The method comprises: Obtain pollutant data; Performing data anomaly processing on the pollutant data to obtain cleaned data; Performing feature conversion and time series data noise reduction on the cleaned data to obtain feature information; Identify the characteristic information through a target pollutant prediction model to obtain pollutant prediction results, including pollutant data prediction results, pollutant category prediction results, and pollutant nonlinear and high-dimensional characteristic data prediction results; Before the step of identifying the characteristic information through a target pollutant prediction model to obtain a pollutant prediction result, the method includes: Optimizing the preset hybrid expert model according to the multi-level knowledge fusion strategy to obtain a preset pollutant prediction model, wherein the preset hybrid expert model includes a linear regression expert model, a support vector regression expert model, a random forest expert model, and a deep neural network expert model; The preset pollutant prediction model is optimized according to the evolutionary learning algorithm strategy to obtain a target pollutant prediction model.

2. The method according to claim 1, wherein The step of optimizing the preset hybrid expert model according to the multi-level knowledge fusion strategy to obtain the preset pollutant prediction model includes: Obtain task requirements for pollutant prediction; According to the task requirements, select the corresponding optimization fusion method in the multi-level knowledge fusion strategy, which includes weighted average fusion, weighted voting fusion and weighted projection fusion; Performing a fusion process on the preset hybrid expert model according to the optimization fusion method to obtain an optimized preset hybrid expert model; The optimized preset hybrid expert model is used as a preset pollutant prediction model.

3. The method according to claim 2, wherein The step of selecting the corresponding optimized fusion method in the multi-level knowledge fusion strategy according to the task requirements includes: When the task requirement is pollutant data prediction, the optimal fusion method is to give priority to weighted average fusion, and the specific expression formula is: in, is the predicted value of pollutants after fusion, For the Pollutant prediction values ​​of expert models, For the The weight of the expert model, and satisfy , is the total number of expert models; When the task requirement is pollutant category prediction, the optimal fusion method is to prioritize weighted voting fusion, and the specific expression formula is: in, is the pollutant prediction category, For the Category of class, is the total number of pollutant categories, For the Expert model prediction category The probability of For the The weight of the expert model; When the task requires the prediction of nonlinear and high-dimensional feature data, the optimal fusion method is to give priority to weighted projection fusion. The specific expression formula is: in, is the predicted value of pollutants after projection, is a high-dimensional vector, represents the projection function, is the optimized output function.

4. The method according to claim 1, wherein The step of optimizing the preset pollutant prediction model according to the evolutionary learning algorithm strategy to obtain the target pollutant prediction model includes: Obtaining hyperparameters and model prediction accuracy corresponding to the expert model of the preset hybrid expert model; Encoding the hyperparameters and weight parameters in the multi-level knowledge fusion strategy to obtain individual gene values; Based on the prediction accuracy of the model, the fitness function is constructed to obtain the fitness value. The fitness function is in the form of: in, is the actual value of fitness, is the predicted value of fitness, is the number of data; Optimizing the fitness value and the individual gene value according to an evolutionary operation strategy and a multi-objective optimization strategy to obtain optimized hyperparameters and optimized fusion weights; The preset pollutant prediction model is updated according to the optimized hyperparameters and the optimized fusion weights to obtain a target pollutant prediction model.

5. The method according to claim 4, wherein The step of optimizing the fitness value and the individual gene value according to the evolutionary operation strategy and the multi-objective optimization strategy to obtain optimized hyperparameters and optimized fusion weights includes: By sorting the fitness values ​​and the individual gene values, selecting individuals corresponding to preset fitness values ​​and preset individual gene values; Performing crossover and mutation operations on the individuals to obtain updated individuals; Optimizing the fitness value of the update individual through a multi-objective optimization strategy to obtain a Pareto optimal solution set; A preset solution is selected from the Pareto optimal solution set and used as the optimized hyperparameter and the optimized fusion weight.

6. The method according to claim 5, wherein The step of optimizing the fitness value of the update individual by a multi-objective optimization strategy to obtain a Pareto optimal solution set includes: Sorting the fitness values ​​of the updated individuals to obtain multiple crowding distance levels; According to the screening of the crowding distance level, the fitness values ​​of multiple updated individuals corresponding to the preset level are selected as Pareto optimal solutions; Based on combining multiple Pareto optimal solutions, a Pareto optimal solution set is obtained.

7. The method according to claim 1, wherein After the step of identifying the characteristic information through the target pollutant prediction model to obtain the pollutant prediction result, the method further includes: Real-time monitoring and acquisition of updated pollutant data; Inputting the updated pollutant data into the target pollutant prediction model to obtain an evaluation performance; When the evaluation performance does not meet the preset prediction requirements, the hyperparameters and fusion weights of the target pollutant prediction model are adjusted until the evaluation performance meets the preset prediction requirements.

8. A pollutant prediction device based on evolutionary learning strategy, characterized in that: The device comprises: Acquisition module, used to obtain pollutant data; A data cleaning module, configured to perform data anomaly processing on the pollutant data to obtain cleaned data; A data conversion module is used to perform feature conversion and time series data noise reduction on the cleaned data to obtain feature information; A result module is used to identify the characteristic information through the target pollutant prediction model to obtain pollutant prediction results, which include pollutant data prediction results, pollutant category prediction results, and pollutant nonlinear and high-dimensional characteristic data prediction results; A construction module is used to optimize the preset hybrid expert model according to a multi-level knowledge fusion strategy to obtain a preset pollutant prediction model. The preset hybrid expert model includes a linear regression expert model, a support vector regression expert model, a random forest expert model and a deep neural network expert model; the preset pollutant prediction model is optimized according to an evolutionary learning algorithm strategy to obtain a target pollutant prediction model.

9. A pollutant prediction device based on evolutionary learning strategy, characterized in that: The device includes: a memory, a processor, and a pollutant prediction program based on an evolutionary learning strategy stored in the memory and running on the processor, wherein the pollutant prediction program based on an evolutionary learning strategy is configured to implement the steps of the pollutant prediction method based on an evolutionary learning strategy as described in any one of claims 1-7.

10. A storage medium, characterized in that: The storage medium stores a pollutant prediction program based on an evolutionary learning strategy. When the pollutant prediction program based on an evolutionary learning strategy is executed by a processor, the steps of the pollutant prediction method based on an evolutionary learning strategy as described in any one of claims 1 to 7 are implemented.

Images