Wetland water quality detection method based on clustering processing

Abstract

The present invention discloses a wetland water quality detection method based on clustering processing, and the method includes data collection, data preprocessing, data preprocessing, clustering processing and wetland water quality detection. The present invention belongs to the field of water quality detection technology, and specifically refers to a wetland water quality detection method based on clustering processing. This scheme adopts a covariance matrix of dynamically updated wetland water quality data to improve the adaptability of wetland detection to environmental changes, and improves the accuracy of detection effect through perturbation analysis and information gain calculation; feature optimization is performed by considering information loss and regularization to improve the convergence of wetland detection; a distance metric based on feature weight is designed by introducing density adjustment, and a similarity function between clusters is designed by introducing a normalization factor and comprehensively comparing all points to achieve accurate division of monitoring points; data is filled based on an LSTM neural network, and cluster evaluation is performed based on a Gaussian mixture model to ensure the accuracy of the final water quality detection results.

Description

Technical Field

[0001] The present invention relates to the technical field of water quality detection, and in particular to a wetland water quality detection method based on clustering processing. Background Art

[0002] Wetland water quality detection methods involve a series of technologies and processes used to monitor and evaluate water quality in wetland environments. These methods are designed to evaluate the physical, chemical, and biological characteristics of water quality in order to understand the health status of wetland ecosystems and environmental changes. However, general wetland water quality detection methods have problems such as low reliability of original data, large amount of calculation due to too many data dimensions, and inability to retain the information of original wetland water quality data during feature screening, which leads to poor wetland water quality detection results; general wetland water quality detection methods have problems such as distance measurement functions that cannot evaluate the differences between water quality detection data, and the inter-cluster similarity of water quality detection data cannot reflect the true differences between water quality detection data clusters, which leads to low accuracy of water quality detection results. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a wetland water quality detection method based on clustering processing. In view of the problems that the general wetland water quality detection method has low reliability of original data, too many data dimensions resulting in large amount of calculation, and the inability to retain the information of the original wetland water quality data during feature screening, thus resulting in poor wetland water quality detection effect, this scheme adopts a covariance matrix of dynamically updated wetland water quality data to improve the adaptability of wetland detection to environmental changes, and through disturbance analysis and information gain calculation, feature screening is made more flexible and adaptable to actual data changes, thereby improving the accuracy of detection effect; feature optimization is performed by considering information loss and regularization , improve the convergence of wetland detection; for the general wetland water quality detection method, the distance measurement function cannot evaluate the differences between water quality detection data, and the inter-cluster similarity of water quality detection data cannot reflect the real differences between water quality detection data clusters, which leads to the low accuracy of water quality detection results. This scheme introduces density adjustment to design a distance measurement based on feature weights, introduces a normalization factor and comprehensively compares all points to design an inter-cluster similarity function, and realizes accurate division of monitoring points, which is suitable for complex water quality conditions; fills the data based on the LSTM neural network, and performs clustering evaluation based on the Gaussian mixture model to ensure the accuracy of the final water quality detection results.

[0004] The technical solution adopted by the present invention is as follows: The wetland water quality detection method based on clustering processing provided by the present invention comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: dimensionality reduction processing;

[0008] Step S4: clustering processing;

[0009] Step S5: wetland water quality detection.

[0010] Furthermore, in step S1, the data collection is the collection of historical wetland monitoring data and real-time wetland monitoring data; both the historical wetland monitoring data and the real-time wetland monitoring data include monitoring data, time series data, environmental parameters, interference source data and spatial dimension data; the historical wetland monitoring data also includes water quality assessment level; the water quality assessment level is used as a data label, which is only used for cluster label selection and does not participate in data processing.

[0011] Furthermore, in step S2, the data preprocessing is to perform data cleaning, data conversion and standardization on the collected data; the data cleaning is to process missing values, duplicate values ​​and outliers on the collected data; the data conversion is to convert the data into a vector form; and the standardization is to perform data standardization based on the maximum and minimum normalization method.

[0012] Furthermore, in step S3, the dimensionality reduction process specifically includes the following steps:

[0013] Step S31: Calculate uncertainty; construct the overall uncertainty of the monitoring data based on the covariance matrix and the sensitivity matrix, expressed as: ; where U is the overall uncertainty of the monitoring data; T is the matrix transpose; S is the sensitivity matrix, which represents the sensitivity of the data features to the overall uncertainty, and the matrix elements , is the jth data feature, U i is the overall uncertainty of the i-th data feature; COV is the covariance matrix of the monitoring data, which represents the covariance between data features. ;COV t+1 and COV t are the covariance matrices at time t+1 and time t, respectively. The covariance matrix is ​​updated every time new monitoring data is collected; N is the number of features; x k is the value of the kth feature; is the characteristic mean; α is the smoothing factor;

[0014] Step S32: Perturbation screening: by downwardly perturbing the relative standard deviation of each feature, observing the change in target parameter uncertainty after the disturbance; expressed as: ; By calculating the contribution of each feature's perturbation to the information gain, important features can be screened; expressed as: ; Calculate feature importance based on random forest, expressed as: ;in, is the change in overall uncertainty caused by the perturbation feature; U(·) is the overall uncertainty; D and are the monitoring data before and after disturbance respectively; IG(·) is the information gain; H(·) is the entropy; is the conditional entropy; is the importance of the jth feature; It is the contribution of the jth feature obtained after training based on the random forest model;

[0015] Step S33: Calculate the disturbance contribution; define the disturbance contribution of each data feature as the ratio of the overall uncertainty change caused by the disturbance to the overall uncertainty, and introduce the minimum disturbance contribution factor to screen the features. The formula used is as follows:

[0016] ;

[0017] ;

[0018] In the formula, is the disturbance contribution factor; and is the weight factor; is the minimum information gain threshold; is the minimum feature importance threshold; Y is the selected feature subset; is the minimum disturbance contribution factor;

[0019] Step S34: feature optimization; establish a search space based on the smoothing factor, weight factor, minimum information gain threshold, minimum feature importance threshold and minimum disturbance contribution factor; introduce a particle swarm optimization algorithm, where each particle represents a combination of a feature subset, using binary coding to implement, and define individual fitness based on minimization of information loss, expressed as: ; Where F is the individual fitness value; is a regularization parameter; based on the particle swarm search algorithm, the optimal parameter setting is found, and a fitness threshold is set in advance. When the individual fitness value is higher than the fitness threshold, the search ends and the optimal feature subset, i.e., the optimized data set, is obtained.

[0020] Furthermore, in step S4, the clustering process specifically includes the following steps:

[0021] Step S41: Initialization; randomly select the initial cluster center; regard each data point as a monitoring point; define the distance metric function as ;in, is the distance metric function; n is the total number of dimensions after dimensionality reduction; k is the dimension index; is the weight of the kth feature; and They are the i1th monitoring point and the j1th monitoring point respectively; and are the kth eigenvalues ​​of the i1th monitoring point and the j1th monitoring point respectively; and are the densities around the i1th monitoring point and the j1th monitoring point respectively;

[0022] Step S42: cluster merging; data allocation to monitoring points based on the metric function; defining a similarity metric function, expressed as ; In each iteration, the similarity between clusters is calculated based on the similarity measurement function, and the two clusters with the largest similarity are merged; where D(·) is the similarity between clusters; C i2 and C j2 are the i2th and j2th clusters respectively; N1 and N2 are clusters C i2 and C j2 The number of monitoring points, k1 and l1 are the corresponding indexes; is the k1th monitoring point in the i2th cluster; is the l1th monitoring point in the j2th cluster;

[0023] Step S43: Data optimization: Introduce LSTM neural network to process time series data to achieve data optimization; BiLSTM completion formula is expressed as: ; Take the data before and after the target segment as input to obtain the optimized data; where, It is to complete the missing time series data segments; It is the data of the first i1 time points before the LSTM network receives the target segment; is the data of the last i1 time points; F(·) is the function of the LSTM neural network, which is used to predict the data; It is a time series segment completed by the LSTM network;

[0024] Step S44: Cluster evaluation; the formula used is: ; Where p is the clustering evaluation value; N3 is the total number of data points, and i3 is the data point index; is to select the maximum value operation, for each data point x i3 , calculate its maximum belonging probability among all clusters; K is the total number of clusters; and are the prior probabilities of the k4th and k5th clusters respectively; and are the mean vectors respectively; and are the covariance matrices of the k4th and k5th clusters respectively; N(·|·, ·) is the Gaussian distribution probability density function; when the clustering evaluation value of this iteration is higher than the evaluation threshold, clustering ends and the clustering result is obtained; if the maximum number of iterations is reached, return to step S41 for re-clustering; otherwise, update and return to step S42 for a new round of iteration.

[0025] Furthermore, in step S5, the wetland water quality detection is based on the clustering results, and the label with the largest number of historical data is used as the cluster label, and the cluster label to which the real-time monitoring data belongs is used as the monitoring result of the real-time monitoring data, that is, the water quality assessment result of the monitoring point, to achieve wetland water quality detection.

[0026] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0027] (1) In view of the problems of low reliability of original data, large amount of calculation due to too many data dimensions, and inability to retain the information of original wetland water quality data during feature screening, which leads to poor wetland water quality detection effect, this scheme adopts the method of dynamically updating the covariance matrix of wetland water quality data to improve the adaptability of wetland detection to environmental changes. Through perturbation analysis and information gain calculation, feature screening is made more flexible and adaptable to actual data changes, thereby improving the accuracy of detection effect. Feature optimization is performed by considering information loss and regularization to improve the convergence of wetland detection.

[0028] (2) In view of the problem that the distance measurement function of general wetland water quality detection methods cannot evaluate the differences between water quality detection data, and the inter-cluster similarity of water quality detection data cannot reflect the real differences between water quality detection data clusters, which leads to low accuracy of water quality detection results, this scheme introduces density adjustment to design a distance measurement based on feature weights, introduces a normalization factor and comprehensively compares all point pairs to design an inter-cluster similarity function, so as to achieve accurate division of monitoring points and is suitable for complex water quality conditions; fills the data based on the LSTM neural network, and performs clustering evaluation based on the Gaussian mixture model to ensure the accuracy of the final water quality detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic diagram of the process of wetland water quality detection method based on clustering processing provided by the present invention;

[0030] Figure 2 is a schematic diagram of the process of step S3;

[0031] Figure 3 It is a schematic diagram of the process of step S4.

[0032] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0034] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0035] Example 1, see Figure 1 The present invention provides a wetland water quality detection method based on clustering processing, which comprises the following steps:

[0036] Step S1: data collection, collecting historical wetland monitoring data and real-time wetland monitoring data;

[0037] Step S2: Data preprocessing, data cleaning, data conversion and standardization of the collected data;

[0038] Step S3: Dimensionality reduction processing, using the covariance matrix of the wetland water quality data to dynamically update, through perturbation analysis and calculation of information gain, considering information loss and regularization to perform feature optimization, and thus complete data dimensionality reduction;

[0039] Step S4: Clustering processing, introducing density adjustment to design a distance metric based on feature weights, designing an inter-cluster similarity function by introducing a normalization factor and comprehensively comparing all point pairs, filling the data based on an LSTM neural network, and performing clustering evaluation based on a Gaussian mixture model to complete the clustering operation;

[0040] Step S5: wetland water quality detection, realizing real-time wetland water quality detection based on clustering results.

[0041] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, both the historical wetland monitoring data and the real-time wetland monitoring data include monitoring data, time series data, environmental parameters, interference source data and spatial dimension data; the historical wetland monitoring data also includes water quality assessment levels; the water quality assessment levels are used as data labels and are only used for cluster label selection and do not participate in data processing.

[0042] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, data cleaning is to process missing values, duplicate values ​​and outliers in the collected data; data conversion is to convert the data into vector form; and standardization is to standardize the data based on the maximum and minimum normalization method.

[0043] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, the dimensionality reduction process specifically includes the following steps:

[0044] Step S31: Calculate uncertainty; construct the overall uncertainty of the monitoring data based on the covariance matrix and the sensitivity matrix, expressed as: ; where U is the overall uncertainty of the monitoring data; T is the matrix transpose; S is the sensitivity matrix, which represents the sensitivity of the data features to the overall uncertainty, and the matrix elements , is the jth data feature, U i is the overall uncertainty of the i-th data feature; COV is the covariance matrix of the monitoring data, which represents the covariance between data features. ;COV t+1 and COV t are the covariance matrices at time t+1 and time t, respectively. The covariance matrix is ​​updated every time new monitoring data is collected; N is the number of features; x k is the value of the kth feature; is the characteristic mean; α is the smoothing factor;

[0045] Step S32: Perturbation screening: by downwardly perturbing the relative standard deviation of each feature, observing the change in target parameter uncertainty after the disturbance; expressed as: ; By calculating the contribution of each feature's perturbation to the information gain, important features can be screened; expressed as: ; Calculate feature importance based on random forest, expressed as: ;in, is the change in overall uncertainty caused by the perturbation feature; U(·) is the overall uncertainty; D and are the monitoring data before and after disturbance respectively; IG(·) is the information gain; H(·) is the entropy; is the conditional entropy; is the importance of the jth feature; It is the contribution of the jth feature obtained after training based on the random forest model;

[0046] Step S33: Calculate the disturbance contribution; define the disturbance contribution of each data feature as the ratio of the overall uncertainty change caused by the disturbance to the overall uncertainty, and introduce the minimum disturbance contribution factor to screen the features. The formula used is as follows:

[0047] ;

[0048] ;

[0049] In the formula, is the disturbance contribution factor; and is the weight factor; is the minimum information gain threshold; is the minimum feature importance threshold; Y is the selected feature subset; is the minimum disturbance contribution factor;

[0050] Step S34: feature optimization; establish a search space based on the smoothing factor, weight factor, minimum information gain threshold, minimum feature importance threshold and minimum disturbance contribution factor; introduce a particle swarm optimization algorithm, where each particle represents a combination of a feature subset, using binary coding to implement, and define individual fitness based on minimization of information loss, expressed as: ; Where F is the individual fitness value; is a regularization parameter; based on the particle swarm search algorithm, the optimal parameter setting is found, and a fitness threshold is set in advance. When the individual fitness value is higher than the fitness threshold, the search ends and the optimal feature subset, i.e., the optimized data set, is obtained.

[0051] By performing the above operations, in view of the problems of low reliability of original data, large amount of calculation caused by too many data dimensions, and inability to retain the information of original wetland water quality data during feature screening, which leads to poor wetland water quality detection effect, this scheme adopts dynamic updating of the covariance matrix of wetland water quality data to improve the adaptability of wetland detection to environmental changes. Through perturbation analysis and information gain calculation, feature screening is made more flexible and adaptable to actual data changes, thereby improving the accuracy of detection effect. Feature optimization is performed by considering information loss and regularization to improve the convergence of wetland detection.

[0052] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, the clustering process specifically includes the following steps:

[0053] Step S41: Initialization; randomly select the initial cluster center; regard each data point as a monitoring point; define the distance metric function as ;in, is the distance metric function; n is the total number of dimensions after dimensionality reduction; k is the dimension index; is the weight of the kth feature; and They are the i1th monitoring point and the j1th monitoring point respectively; and are the kth eigenvalues ​​of the i1th monitoring point and the j1th monitoring point respectively; and are the densities around the i1th monitoring point and the j1th monitoring point respectively;

[0054] Step S42: cluster merging; data allocation to monitoring points based on the metric function; defining a similarity metric function, expressed as ; In each iteration, the similarity between clusters is calculated based on the similarity measurement function, and the two clusters with the largest similarity are merged; where D(·) is the similarity between clusters; C i2 and C j2 are the i2th and j2th clusters respectively; N1 and N2 are clusters C i2 and C j2 The number of monitoring points, k1 and l1 are the corresponding indexes; is the k1th monitoring point in the i2th cluster; is the l1th monitoring point in the j2th cluster;

[0055] Step S43: Data optimization: Introduce LSTM neural network to process time series data to achieve data optimization; BiLSTM completion formula is expressed as: ; Take the data before and after the target segment as input to obtain the optimized data; where, It is to complete the missing time series data segments; It is the data of the first i1 time points before the LSTM network receives the target segment; is the data of the last i1 time points; F(·) is the function of the LSTM neural network, which is used to predict the data; It is a time series segment completed by the LSTM network;

[0056] Step S44: Cluster evaluation; the formula used is: ; Where p is the clustering evaluation value; N3 is the total number of data points, and i3 is the data point index; is to select the maximum value operation, for each data point x i3, calculate its maximum belonging probability among all clusters; K is the total number of clusters; and are the prior probabilities of the k4th and k5th clusters respectively; and are the mean vectors respectively; and are the covariance matrices of the k4th and k5th clusters respectively; N(·|·, ·) is the Gaussian distribution probability density function; when the clustering evaluation value of this iteration is higher than the evaluation threshold, clustering ends and the clustering result is obtained; if the maximum number of iterations is reached, return to step S41 for re-clustering; otherwise, update and return to step S42 for a new round of iteration.

[0057] By performing the above operations, in view of the problems of low reliability of original data, large amount of calculation caused by too many data dimensions, and inability to retain the information of original wetland water quality data during feature screening, which leads to poor wetland water quality detection effect, this scheme adopts dynamic updating of the covariance matrix of wetland water quality data to improve the adaptability of wetland detection to environmental changes. Through perturbation analysis and information gain calculation, feature screening is made more flexible and adaptable to actual data changes, thereby improving the accuracy of detection effect. Feature optimization is performed by considering information loss and regularization to improve the convergence of wetland detection.

[0058] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, wetland water quality detection is based on clustering results. The label with the largest number of historical data is used as the cluster label, and the cluster label to which the real-time monitoring data belongs is used as the monitoring result of the real-time monitoring data, that is, the water quality assessment result of the monitoring point, so as to realize wetland water quality detection.

[0059] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0060] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0061] The present invention and its embodiments are described above, which is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it and design structural methods and embodiments similar to the technical solution without creativity without departing from the purpose of the invention, they should all fall within the protection scope of the present invention.

Claims

1. A wetland water quality detection method based on clustering processing, characterized in that: The method comprises the following steps: Step S1: data collection, collecting historical wetland monitoring data and real-time wetland monitoring data; Step S2: data preprocessing; Step S3: dimensionality reduction processing; Step S4: clustering processing; Step S5: wetland water quality detection, realizing real-time wetland water quality detection based on clustering results; In step S3, the dimensionality reduction process includes the following steps: Step S31: Calculate uncertainty; construct the overall uncertainty of the monitoring data based on the covariance matrix and the sensitivity matrix, expressed as: ; where U is the overall uncertainty of the monitoring data; T is the matrix transpose; S is the sensitivity matrix, which represents the sensitivity of the data features to the overall uncertainty, and the matrix elements , is the jth data feature, U i is the overall uncertainty of the i-th data feature; COV is the covariance matrix of the monitoring data, which represents the covariance between data features. ;COV t+1 and COV t are the covariance matrices at time t+1 and time t, respectively. The covariance matrix is ​​updated every time new monitoring data is collected; N is the number of features; x k is the value of the kth feature; is the characteristic mean; α is the smoothing factor; Step S32: Perturbation screening: by downwardly perturbing the relative standard deviation of each feature, observing the change in target parameter uncertainty after the disturbance; expressed as: ; By calculating the contribution of each feature's perturbation to the information gain, important features can be screened; expressed as: ; Calculate feature importance based on random forest, expressed as: ;in, is the change in overall uncertainty caused by the perturbation feature; U(·) is the overall uncertainty; D and are the monitoring data before and after disturbance respectively; IG(·) is the information gain; H(·) is the entropy; is the conditional entropy; is the importance of the jth feature; It is the contribution of the jth feature obtained after training based on the random forest model; Step S33: Calculate the disturbance contribution; define the disturbance contribution of each data feature as the ratio of the overall uncertainty change caused by the disturbance to the overall uncertainty, and introduce the minimum disturbance contribution factor to screen the features. The formula used is as follows: ； ； In the formula, is the disturbance contribution factor; and is the weight factor; is the minimum information gain threshold; is the minimum feature importance threshold; Y is the selected feature subset; is the minimum perturbation contribution factor; IG(·) is the information gain; is the jth data feature; is the importance of the jth feature; Step S34: feature optimization; establish a search space based on the smoothing factor, weight factor, minimum information gain threshold, minimum feature importance threshold and minimum disturbance contribution factor; introduce a particle swarm optimization algorithm, where each particle represents a combination of a feature subset, using binary coding to implement, and define individual fitness based on minimization of information loss, expressed as: ; Where F is the individual fitness value; is a regularization parameter; based on the particle swarm search algorithm, the optimal parameter setting is found, and a fitness threshold is set in advance. When the individual fitness value is higher than the fitness threshold, the search ends and the optimal feature subset, i.e., the optimized data set, is obtained.

2. The wetland water quality detection method based on clustering processing according to claim 1 is characterized in that: In step S4, the clustering process specifically includes the following steps: Step S41: Initialization; randomly select the initial cluster center; regard each data point as a monitoring point; define the distance metric function as ;in, is the distance metric function; n is the total number of dimensions after dimensionality reduction; k is the dimension index; is the weight of the kth feature; and They are the i1th monitoring point and the j1th monitoring point respectively; and are the kth eigenvalues ​​of the i1th monitoring point and the j1th monitoring point respectively; and are the densities around the i1th monitoring point and the j1th monitoring point respectively; Step S42: cluster merging; data allocation to monitoring points based on the metric function; defining a similarity metric function, expressed as ; In each iteration, the similarity between clusters is calculated based on the similarity measurement function, and the two clusters with the largest similarity are merged; where D(·) is the similarity between clusters; C i2 and C j2 are the i2th and j2th clusters respectively; N1 and N2 are clusters C i2 and C j2 The number of monitoring points, k1 and l1 are the corresponding indexes; is the k1th monitoring point in the i2th cluster; is the l1th monitoring point in the j2th cluster; Step S43: Data optimization: Introduce LSTM neural network to process time series data to achieve data optimization; BiLSTM completion formula is expressed as: ; Take the data before and after the target segment as input to obtain the optimized data; where, It is to complete the missing time series data segments; It is the data of the first i1 time points before the LSTM network receives the target segment; is the data of the last i1 time points; F(·) is the function of the LSTM neural network, which is used to predict the data; It is a time series segment completed by the LSTM network; Step S44: Cluster evaluation; the formula used is: ; Where p is the clustering evaluation value; N3 is the total number of data points, and i3 is the data point index; is to select the maximum value operation, for each data point x i3 , calculate its maximum belonging probability among all clusters; K is the total number of clusters; and are the prior probabilities of the k4th and k5th clusters respectively; and are the mean vectors respectively; and are the covariance matrices of the k4th and k5th clusters respectively; N(·|·, ·) is the Gaussian distribution probability density function; when the clustering evaluation value of this iteration is higher than the evaluation threshold, clustering ends and the clustering result is obtained; if the maximum number of iterations is reached, return to step S41 for re-clustering; otherwise, update and return to step S42 for a new round of iteration.

3. The wetland water quality detection method based on clustering processing according to claim 1 is characterized in that: In step S1, the historical wetland monitoring data and the real-time wetland monitoring data both include monitoring data, time series data, environmental parameters, interference source data and spatial dimension data; the historical wetland monitoring data also includes water quality assessment levels; the water quality assessment levels are used as data labels, which are only used for cluster label selection and do not participate in data processing.

4. The wetland water quality detection method based on clustering processing according to claim 1 is characterized in that: In step S5, the wetland water quality detection is based on the clustering results, and the label with the largest number of historical data is used as the cluster label, and the cluster label to which the real-time monitoring data belongs is used as the monitoring result of the real-time monitoring data, that is, the water quality assessment result of the monitoring point, to achieve wetland water quality detection.

5. The wetland water quality detection method based on clustering processing according to claim 1 is characterized in that: In step S2, the data preprocessing is to perform data cleaning, data conversion and standardization on the collected data; the data cleaning is to process missing values, duplicate values ​​and outliers on the collected data; the data conversion is to convert the data into a vector form; and the standardization is to standardize the data based on the maximum and minimum normalization method.

Images