A data clustering ensemble fusion method and device

Abstract

The application discloses a data clustering integrated fusion method and device ， The application relates to the technical field of data processing ； The method classifies the environment monitoring data of a target region through multiple preset clustering algorithms to obtain data cluster sets corresponding to the clustering algorithms; the data clusters in each data cluster set are grouped according to similarity to obtain data cluster groups; for each data cluster group, common data and non-common data in the data cluster group are determined; the data characteristics of the data cluster group are determined according to the common data, effective data in the non-common data that meets the data characteristics is determined, and the effective data and the common data are combined to obtain a data clustering result corresponding to the data cluster group; and the data clustering result corresponding to each data cluster group is determined to obtain an integrated data clustering result, so that low-quality clustering members in each data cluster are reduced, negative effects of the low-quality basic clustering members on the final integrated fusion result are avoided, and the overall analysis accuracy and reliability are improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically relating to a data clustering, integration, and fusion method and apparatus. Background Technology

[0002] In order to comprehensively and accurately assess environmental quality, multi-dimensional environmental testing is usually conducted, such as air quality monitoring, water quality testing, and noise pollution measurement.

[0003] These detection methods generate a large amount of data, each set reflecting the state of the environment in a specific aspect. To better understand and analyze this data, and to uncover the environmental patterns and trends behind it, it is necessary to perform clustering, integration, and fusion processing on this data.

[0004] Clustering ensemble processing is an advanced data analysis technique that integrates data from different detection methods and time periods. Through clustering algorithms, it identifies similarities and differences within the data, forming meaningful data clusters. These clusters can help us gain a clearer understanding of the overall environmental situation and uncover potential environmental problems.

[0005] However, existing data clustering ensemble processing methods usually treat all clusters and basic clustering algorithms equally, resulting in low-quality basic cluster members. These low-quality basic cluster members may negatively affect the final ensemble results, reducing the overall accuracy and reliability of the analysis. Summary of the Invention

[0006] The purpose of this invention is to address the problem that low-quality basic cluster members may negatively impact the final integration and fusion results, reducing the overall accuracy and reliability of the analysis. Therefore, this invention proposes a data clustering integration and fusion method and apparatus.

[0007] In a first aspect of this invention, a data clustering integration and fusion method is first proposed, the method comprising: Environmental monitoring data of the target area are classified using multiple preset clustering algorithms to obtain a data cluster set corresponding to each clustering algorithm; the number of data clusters set for each clustering algorithm is the same; The data clusters in each data cluster set are grouped according to similarity to obtain data cluster groups; the number of data cluster groups is the same as the number of data clusters; each data cluster group contains one data cluster from each data cluster set. For each data cluster, determine the common data and non-common data within the data cluster; common data is data contained in all data clusters within the data cluster; non-common data is data in the data cluster other than the common data. Based on the shared data, determine the data characteristics of the data cluster group, identify the valid data in the non-shared data that meets the data characteristics, and combine the valid data and the shared data to obtain the data clustering result corresponding to the data cluster group. The data clustering results corresponding to each data cluster group are determined to obtain the integrated data clustering results. The environmental pollution level corresponding to the target area is obtained based on the integrated data clustering results.

[0008] Optionally, data clusters in each data cluster set can be grouped according to similarity to obtain data cluster groups, including: For each data cluster in the data cluster set, the average value of the coordinates corresponding to each data in the data cluster is calculated to obtain the first target coordinates; The second target coordinates are obtained by averaging the centroid coordinates of the data cluster with the first target coordinates; these centroid coordinates are generated by the clustering algorithm when classifying environmental monitoring data. Obtain any data cluster set, and group the data clusters within that data cluster set to obtain target data groups; each target data group corresponds to a second target coordinate. For each remaining data cluster, the target distance between the second target coordinates corresponding to each data cluster in the data cluster and the second target coordinates corresponding to the target data group is determined sequentially. Based on the minimum target distance, the data clusters in the data cluster are grouped to obtain data cluster groups.

[0009] Optionally, the data characteristics of the data cluster group are determined based on the shared data, the valid data in the non-shared data that meets the data characteristics are identified, and the valid data and shared data are combined to obtain the data clustering result corresponding to the data cluster group, including: Based on this shared data, fit a standard multivariate normal distribution function; Calculate the expected value of each data point in the non-public data set falling on the multivariate normal distribution function; Data that is expected to be greater than the preset value is recorded as valid data. The effective data and the shared data are combined to obtain the data clustering result corresponding to the data cluster group.

[0010] Optionally, after calculating the expected value of each data point in the non-public data falling on the multivariate normal distribution function, the following steps are also included: Data that is less than the preset value is recorded as invalid data. The data cluster corresponding to the invalid data is recorded as the original data cluster. For the original data cluster, the replacement data cluster that is closest to the original data cluster is obtained through the second target coordinates corresponding to the original data cluster. The replacement data cluster is the data cluster that is adjacent to the original data cluster. The invalid data is assigned to the replacement data cluster, and the target expected value of the invalid data falling on the multivariate normal distribution function corresponding to the replacement data cluster is calculated. If the target expectation is greater than the preset target expectation value, then the invalid data is classified into the replacement data cluster.

[0011] Optionally, classifying the invalid data into a replacement data cluster and calculating the target expected value of the invalid data falling on the multivariate normal distribution function corresponding to the replacement data cluster further includes: Step 1: If the target expectation is less than the preset target expectation value, then the invalid data is assigned to the next nearest replacement data cluster; Step 2: Recalculate the target expected value of the invalid data falling on the multivariate normal distribution function corresponding to the replacement data cluster; Step 3: If the target expectation is still less than the preset target expectation value, repeat Step 1 and Step 2 above until the target expectation of the invalid data for all replacement data clusters is less than the preset target expectation value. Then, record the invalid data as special data and classify it separately.

[0012] In a second aspect of this invention, a data clustering integration and fusion device is proposed, comprising: a data clustering module, a data cluster group module, a shared data module, a data clustering result module, and an environmental pollution level determination module. The data clustering module is used to classify environmental monitoring data of the target area using multiple preset clustering algorithms to obtain a data clustering set corresponding to each clustering algorithm; the number of data clusters set for each clustering algorithm is the same; The data cluster grouping module is used to group data clusters in each data cluster set according to similarity to obtain data cluster groups; the number of data cluster groups is the same as the number of data clusters; each data cluster group contains one data cluster from each data cluster set; The shared data module is used to determine the shared data and non-shared data in each data cluster group; shared data is data contained in all data clusters in the data cluster group; non-shared data is data in the data cluster group other than shared data. The data clustering result module is used to determine the data characteristics of the data cluster group based on the common data, determine the valid data in the non-common data that meets the data characteristics, and combine the valid data and the common data to obtain the data clustering result corresponding to the data cluster group. The environmental pollution level determination module is used to determine the data clustering results corresponding to each data cluster group to obtain the integrated data clustering results, and to obtain the environmental pollution level corresponding to the target area based on the integrated data clustering results.

[0013] Optionally, the data cluster group module includes a first target coordinate module, a second target coordinate module, a target data group module, and a data cluster grouping module: The first target coordinate module is used to calculate the average value of the coordinates of each data in each data cluster to obtain the first target coordinates for each data cluster in each data cluster set; The second target coordinate module is used to average the centroid coordinates of the data cluster with the first target coordinates to obtain the second target coordinates; the centroid coordinates are generated by the clustering algorithm when classifying environmental monitoring data; The target data group module is used to acquire any data cluster set, and group the data clusters within the data cluster set to obtain target data groups; each target data group corresponds to a second target coordinate. The data cluster grouping module is used to determine the target distance between the second target coordinates corresponding to each data cluster in the data cluster set and the second target coordinates corresponding to the target data group for each remaining data cluster set. The data clusters in the data cluster set are grouped based on the minimum target distance value to obtain a data cluster group.

[0014] Optionally, the data clustering result module includes a multivariate normal distribution function module, an expected distribution module, an effective data module, and a data combination module: The multivariate normal distribution function module is used to fit a standard multivariate normal distribution function based on the shared data; The expected distribution module is used to calculate the expected value of each data point in the non-public data falling on the multivariate normal distribution function; The valid data module is used to record data that is expected to be greater than a preset value as valid data; The data combining module is used to combine the valid data and the shared data to obtain the data clustering result corresponding to the data cluster group.

[0015] Optionally, the expected distribution module includes an invalid data module, a replacement data cluster module, a target expected module, and a data classification module: The invalid data module is used to record data that is less than a preset value as invalid data; The replacement data cluster module is used to record the data cluster corresponding to the invalid data as the original data cluster, and for the original data cluster, to obtain the replacement data cluster that is closest to the original data cluster through the second target coordinates corresponding to the original data cluster; the replacement data cluster is the data cluster adjacent to the original data cluster. The target expectation module is used to classify the invalid data into a replacement data cluster and calculate the target expectation that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster. The data classification module is used to classify invalid data into the replacement data cluster if the target expectation is greater than the preset target expectation value.

[0016] Optionally, the target expectation module includes an invalid data partitioning module, a recalculation module, and a separate classification module: The invalid data partitioning module is used to partition the invalid data into the next nearest replacement data cluster if the target expectation is less than the preset target expectation value. The recalculation module is used to recalculate the target expectation that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster; The separate classification module is used to repeat the replacement data cluster module and the recalculation module as described above if the target expectation is still less than the preset target expectation value, until the target expectation of the invalid data is less than the preset target expectation value for all replacement data clusters, and then the invalid data is recorded as special data and classified separately.

[0017] The beneficial effects of this invention are: This invention proposes a data clustering integration method. It classifies environmental monitoring data of a target area using multiple preset clustering algorithms, obtaining data cluster sets corresponding to each algorithm. The data clusters in each set are grouped according to similarity to form data cluster groups. For each data cluster group, common and non-common data are identified. Data characteristics of the data cluster group are determined based on the common data, and valid data satisfying these characteristics in the non-common data are identified. The valid data and common data are combined to obtain the data clustering result corresponding to the data cluster group. The integrated data clustering result is obtained by determining the data clustering result for each data cluster group, and the environmental pollution level of the target area is determined based on the integrated data clustering result. By grouping these data clusters according to the similarity between them, common and non-common data are identified within each data cluster group. Data characteristics of the data cluster group are determined based on the common data, and valid data satisfying these characteristics in the non-common data are identified based on these characteristics. This method reduces the number of low-quality cluster members in each data cluster, avoiding the negative impact of these low-quality basic cluster members on the final integration result, and improving the overall accuracy and reliability of the analysis. Attached Figure Description

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Figure 1 A flowchart of a data clustering integration and fusion method is provided in Embodiment 1 of the present invention; Figure 2 The present invention provides a structural diagram of a data clustering integration and fusion device. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Embodiment 1 of this invention provides a data clustering integration and fusion method. See also... Figure 1 , Figure 1 This is a flowchart illustrating a data clustering integration and fusion method provided in Embodiment 1 of the present invention. The method includes the following steps: S101, the environmental monitoring data of the target area is classified by multiple preset clustering algorithms to obtain the data cluster set corresponding to each clustering algorithm.

[0022] S102, group the data clusters in each data cluster set according to their similarity to obtain data cluster groups.

[0023] S103, for each data cluster group, determine the common data and non-common data in the data cluster group.

[0024] S104. Determine the data characteristics of the data cluster group based on the shared data, identify the valid data in the non-shared data that meets the data characteristics, and combine the valid data and the shared data to obtain the data clustering result corresponding to the data cluster group.

[0025] S105, determine the data clustering results corresponding to each data cluster group to obtain the integrated data clustering results, and obtain the environmental pollution level corresponding to the target area based on the integrated data clustering results.

[0026] In this clustering algorithm, the number of data clusters is the same for each clustering algorithm; the number of data cluster groups is the same as the number of data clusters; each data cluster group contains one data cluster from each data cluster set; common data is the data contained in each data cluster of the data cluster group; non-common data is the data in the data cluster group other than the common data.

[0027] Based on the data clustering integration and fusion method provided by the embodiments of the present invention, these data clusters are grouped by similarity between them to determine common and non-common data in the data cluster group. The data characteristics of the data cluster group are determined based on the common data, and then the effective data in the non-common data that meets the data characteristics are determined based on the data characteristics. This reduces the number of low-quality cluster members in each data cluster, avoids the negative impact of these low-quality basic cluster members on the final integration and fusion result, and improves the overall analysis accuracy and reliability.

[0028] In one implementation, the target area is a city, which is divided into multiple sub-areas. Each sub-area is equipped with an environmental monitoring terminal. The environmental monitoring data includes air quality data, water quality test data, noise pollution measurement, etc. For example, air quality testing equipment is placed in areas such as streetlights, office buildings, and parks; water quality testing equipment is placed in areas such as tap water pipelines and reservoirs; and noise pollution testing equipment is placed in areas such as roads, railways, residential areas, and commercial streets.

[0029] In one implementation, the environmental monitoring data is high-dimensional data, which can be analyzed from dimensions such as time, space, density of human activities, and ecosystem.

[0030] In one implementation, the preset clustering algorithm can be K-means, hierarchical clustering, DBSCAN, etc. Using multiple preset clustering algorithms can enhance the robustness and effectiveness of clustering, improve the processing of high-dimensional data, and reduce the impact of invalid data on the clustering results.

[0031] In one implementation, grouping data clusters based on their similarity can more accurately identify data features.

[0032] In one implementation, shared data is used to determine data characteristics, which further improves the accuracy of clustering and avoids the negative impact of low-quality basic cluster members on the final integration result, thereby improving the overall accuracy and reliability of the analysis.

[0033] In one implementation, each data cluster can belong to only one data cluster group.

[0034] In Example 2, step S102 includes the following steps: Step 1: For each data cluster in the data cluster set, calculate the average value of the coordinates corresponding to each data point in the data cluster to obtain the first target coordinates.

[0035] Step 2: Average the centroid coordinates of the data cluster with the first target coordinates to obtain the second target coordinates.

[0036] Step 3: Obtain any data cluster set, and group the data clusters within the data cluster set to obtain the target data group; each target data group corresponds to a second target coordinate.

[0037] Step 4: For each remaining data cluster, sequentially determine the target distance between the second target coordinates corresponding to each data cluster in the data cluster and the second target coordinates corresponding to the target data group. Using the minimum target distance as a condition, group the data clusters in the data cluster to obtain data cluster groups.

[0038] The centroid coordinates are generated by the clustering algorithm when classifying environmental monitoring data; In one implementation, the data clusters within the data set are grouped based on the minimum target distance, resulting in data cluster groups. Taking a data cluster group with three data clusters and corresponding to three datasets as an example, the first group of data clusters is denoted as a11, a12, a13; the second group as a21, a22, a23; and the third group as a31, a32, a33. If the distance between a11 and a21 is 0.1, and the distance between a11 and a22 is 0.5... The distance between a11 and a23 is 0.8, the distance between a11 and a31 is 0.11, the distance between a11 and a32 is 0.57, the distance between a11 and a33 is 0.78, the distance between a12 and a21 is 0.5, the distance between a12 and a22 is 0.08, the distance between a12 and a23 is 0.3, the distance between a12 and a31 is 0.51, the distance between a12 and a32 is 0.07, and the distance between a12 and a33 is... The distance between a13 and a21 is 0.28, the distance between a13 and a22 is 0.3, the distance between a13 and a23 is 0.11, the distance between a13 and a31 is 0.82, the distance between a11 and a32 is 0.32, the distance between a11 and a33 is 0.08, the distance between a21 and a31 is 0.09, the distance between a21 and a32 is 0.47, and the distance between a21 and a33 is 0.75. The distance between a22 and a31 is 0.44, the distance between a22 and a32 is 0.13, the distance between a22 and a33 is 0.34, the distance between a23 and a31 is 0.84, the distance between a23 and a32 is 0.33, and the distance between a23 and a33 is 0.08. Therefore, a11, a21, and a31 are denoted as one data cluster, a12, a22, and a32 are denoted as another data cluster, and a13, a23, and a33 are denoted as yet another data cluster.

[0039] In one implementation, if the clustering algorithm is K-means clustering, then the centroid coordinates are the average of all points in the cluster.

[0040] In Example 3, step S104 includes the following steps: Step 1: Based on the shared data, fit a standard multivariate normal distribution function.

[0041] Step 2: Calculate the expected value of each data point in the non-public data falling on the multivariate normal distribution function.

[0042] Step 3: Record data that is expected to be greater than the preset value as valid data.

[0043] Step 4: Combine the valid data and the shared data to obtain the data clustering result corresponding to the data cluster group.

[0044] In one implementation, the preset value is the expected minimum value that falls on the multivariate normal distribution function.

[0045] In one implementation, the validity of the data can be more accurately assessed by fitting a standard multivariate normal distribution function and calculating the expected value of non-public data falling on it based on the function.

[0046] In Example 4, after step two in Example 3, the following is also included: Step 5: Record data that is less than the preset value as invalid data.

[0047] Step 6: Record the data cluster corresponding to the invalid data as the original data cluster. For the original data cluster, obtain the replacement data cluster that is closest to the original data cluster through the second target coordinates corresponding to the original data cluster.

[0048] Step 7: Assign the invalid data to the replacement data cluster and calculate the target expected value of the invalid data falling on the multivariate normal distribution function corresponding to the replacement data cluster.

[0049] Step 8: If the target expectation is greater than the preset target expectation value, then the invalid data is classified into the replacement data cluster.

[0050] The replacement data cluster is the data cluster adjacent to the original data cluster.

[0051] In one implementation, invalid data can be effectively prevented from affecting the current data cluster by allocating it to a replacement data cluster. This ensures data accuracy and reliability, avoiding situations where invalid data is invalid in the current data cluster but might be valid in the original data cluster.

[0052] In Example 5, step seven of Example 4 further includes: 1. If the target expectation is less than the preset target expectation value, then the invalid data will be assigned to the next nearest replacement data cluster; 2. Recalculate the target expected value that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster; 3. If the target expectation is still less than the preset target expectation value, repeat steps 1 and 2 above until the target expectation value of the invalid data for all replacement data clusters is less than the preset target expectation value. Then, record the invalid data as special data and classify it separately.

[0053] In one implementation, the invalid data is determined to have a categorized data cluster by calculating the target expectation that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster, thereby reducing the risk of misjudgment.

[0054] In one implementation, invalid data is recorded as special data and classified separately to avoid the negative impact of invalid data on other data clusters, thereby improving the overall accuracy and reliability of the analysis.

[0055] Based on the same inventive concept, embodiments of the present invention also provide a data clustering, integration, and fusion device. See also Figure 2 , Figure 2 A structural diagram of a data clustering integration and fusion device provided in an embodiment of the present invention includes: The data clustering module is used to classify environmental monitoring data of the target area using multiple preset clustering algorithms to obtain the data clusters corresponding to each clustering algorithm; the number of data clusters set for each clustering algorithm is the same; The data cluster group module is used to group data clusters in various data cluster sets according to similarity to obtain data cluster groups; the number of data cluster groups is the same as the number of data clusters; each data cluster group contains one data cluster from each data cluster set; The shared data module is used to determine the shared data and non-shared data in each data cluster group; shared data is data contained in all data clusters in the data cluster group; non-shared data is data in the data cluster group other than shared data. The data clustering result module is used to determine the data characteristics of the data cluster group based on the common data, determine the valid data in the non-common data that meets the data characteristics, and combine the valid data and the common data to obtain the data clustering result corresponding to the data cluster group; The environmental pollution level determination module is used to determine the data clustering results corresponding to each data cluster group to obtain the integrated data clustering results, and to obtain the environmental pollution level corresponding to the target area based on the integrated data clustering results.

[0056] Based on the data clustering integration and fusion device provided in this embodiment of the invention, the data clusters are grouped by similarity between them to determine common and non-common data in the data cluster group. The data characteristics of the data cluster group are determined based on the common data, and then the valid data in the non-common data that meets the data characteristics are determined based on the data characteristics. This reduces the number of low-quality cluster members in each data cluster, avoids the negative impact of these low-quality basic cluster members on the final integration and fusion result, and improves the overall analysis accuracy and reliability.

[0057] In one embodiment, the data cluster module includes: The first target coordinate module is used to calculate the average value of the coordinates of each data in each data cluster to obtain the first target coordinates for each data cluster in each data cluster set; The second target coordinate module is used to average the centroid coordinates of the data cluster with the first target coordinates to obtain the second target coordinates; these centroid coordinates are generated by the clustering algorithm when classifying environmental monitoring data. The target data group module is used to obtain any data cluster set, and group the data clusters within that data cluster set to obtain target data groups; each target data group corresponds to a second target coordinate. The data cluster grouping module is used to determine the target distance between the second target coordinates of each data cluster in the remaining data cluster set and the second target coordinates of the target data group, and to group the data clusters in the data cluster set based on the minimum target distance value, so as to obtain the data cluster group.

[0058] In one embodiment, the data clustering result module includes: The multivariate normal distribution function module is used to fit a standard multivariate normal distribution function based on the shared data; The Expected Distribution module is used to calculate the expected value of each data point in the non-public data falling on the multivariate normal distribution function; The valid data module is used to record data that is expected to be greater than a preset value as valid data. The data combination module is used to combine the valid data and the shared data to obtain the data clustering result corresponding to the data cluster group.

[0059] In one embodiment, the desired distribution module includes: The invalid data module is used to record data that is less than the preset value as invalid data. The data cluster replacement module is used to record the data cluster corresponding to the invalid data as the original data cluster. For the original data cluster, the replacement data cluster that is closest to the original data cluster is obtained through the second target coordinates corresponding to the original data cluster. The replacement data cluster is the data cluster that is adjacent to the original data cluster. The target expectation module is used to classify the invalid data into the replacement data cluster and calculate the target expectation that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster. The data classification module is used to classify invalid data into a replacement data cluster if the target expectation is greater than the preset target expectation value.

[0060] In one embodiment, the target expectation module includes: The invalid data partitioning module is used to partition invalid data into the next nearest replacement data cluster if the target expectation is less than the preset target expectation value; The recalculation module is used to recalculate the target expectation that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster; The separate classification module is used to repeat the above-mentioned data cluster replacement module and recalculation module if the target expectation is still less than the preset target expectation value, until the target expectation of the invalid data is less than the preset target expectation value for all replacement data clusters. Then, the invalid data is recorded as special data and classified separately.

[0061] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A data clustering integration and fusion method, characterized in that, The method includes: Environmental monitoring data of the target area are classified using multiple preset clustering algorithms to obtain a data cluster set corresponding to each clustering algorithm; the number of data clusters set for each clustering algorithm is the same. Data clusters in each data cluster set are grouped according to similarity to obtain data cluster groups; The number of data cluster groups is the same as the number of data clusters; each data cluster group contains one data cluster from each data cluster set. For each data cluster, determine the common data and non-common data within the data cluster; common data is data contained in all data clusters within the data cluster; non-common data is data in the data cluster other than the common data. Based on the shared data, determine the data characteristics of the data cluster group, identify the valid data in the non-shared data that meets the data characteristics, and combine the valid data and the shared data to obtain the data clustering result corresponding to the data cluster group. The data clustering results corresponding to each data cluster group are determined to obtain the integrated data clustering results. The environmental pollution level corresponding to the target area is obtained based on the integrated data clustering results. Based on the shared data, the data characteristics of the data cluster group are determined. Then, valid data that meets these characteristics is identified from the non-shared data. Combining this valid data with the shared data yields the data clustering results corresponding to the data cluster group, including: Based on this shared data, fit a standard multivariate normal distribution function; Calculate the expected value of each data point in the non-public data set falling on the multivariate normal distribution function; Data that is expected to be greater than the preset value is recorded as valid data. The effective data and the shared data are combined to obtain the data clustering result corresponding to the data cluster group.

2. The data clustering integration and fusion method according to claim 1, characterized in that, The data clusters in each data cluster set are grouped according to similarity to obtain data cluster groups, including: For each data cluster in the data cluster set, the average value of the coordinates corresponding to each data in the data cluster is calculated to obtain the first target coordinates; The second target coordinates are obtained by averaging the centroid coordinates of the data cluster with the first target coordinates; these centroid coordinates are generated by the clustering algorithm when classifying environmental monitoring data. Obtain any data cluster set, and group the data clusters within that data cluster set to obtain target data groups; each target data group corresponds to a second target coordinate. For each remaining data cluster, the target distance between the second target coordinates corresponding to each data cluster in the data cluster and the second target coordinates corresponding to the target data group is determined sequentially. Based on the minimum target distance, the data clusters in the data cluster are grouped to obtain data cluster groups.

3. The data clustering integration and fusion method according to claim 1, characterized in that, After calculating the expected value of each data point in the non-public data falling on the multivariate normal distribution function, the following steps are also included: Data that is less than the preset value is recorded as invalid data. The data cluster corresponding to the invalid data is recorded as the original data cluster. For the original data cluster, the replacement data cluster that is closest to the original data cluster is obtained through the second target coordinates corresponding to the original data cluster. The replacement data cluster is the data cluster that is adjacent to the original data cluster. The invalid data is assigned to the replacement data cluster, and the target expected value of the invalid data falling on the multivariate normal distribution function corresponding to the replacement data cluster is calculated. If the target expectation is greater than the preset target expectation value, then the invalid data is classified into the replacement data cluster.

4. The data clustering integration and fusion method according to claim 3, characterized in that, The process of assigning the invalid data to the replacement data cluster and calculating the expected value of the invalid data falling on the multivariate normal distribution function corresponding to the replacement data cluster also includes: Step 1: If the target expectation is less than the preset target expectation value, then the invalid data is assigned to the next nearest replacement data cluster; Step 2: Recalculate the target expected value of the invalid data falling on the multivariate normal distribution function corresponding to the replacement data cluster; Step 3: If the target expectation is still less than the preset target expectation value, repeat Step 1 and Step 2 above until the target expectation of the invalid data for all replacement data clusters is less than the preset target expectation value. Then, record the invalid data as special data and classify it separately.

5. A data clustering integration and fusion device, characterized in that, The device includes a data clustering module, a data cluster group module, a shared data module, a data clustering result module, and an environmental pollution level determination module. The data clustering module is used to classify environmental monitoring data of the target area using multiple preset clustering algorithms to obtain a data clustering set corresponding to each clustering algorithm; the number of data clusters set for each clustering algorithm is the same; The data cluster grouping module is used to group data clusters in each data cluster set according to similarity to obtain data cluster groups; The number of data cluster groups is the same as the number of data clusters; each data cluster group contains one data cluster from each data cluster set. The shared data module is used to determine the shared data and non-shared data in each data cluster group; shared data is data contained in all data clusters in the data cluster group; non-shared data is data in the data cluster group other than shared data. The data clustering result module is used to determine the data characteristics of the data cluster group based on the common data, determine the valid data in the non-common data that meets the data characteristics, and combine the valid data and the common data to obtain the data clustering result corresponding to the data cluster group. The environmental pollution level determination module is used to determine the data clustering result corresponding to each data cluster group to obtain the integrated data clustering result, and to obtain the environmental pollution level corresponding to the target area based on the integrated data clustering result; The data clustering result module includes a multivariate normal distribution function module, an expected distribution module, an effective data module, and a data combination module: The multivariate normal distribution function module is used to fit a standard multivariate normal distribution function based on the shared data; The expected distribution module is used to calculate the expected value of each data point in the non-public data falling on the multivariate normal distribution function; The valid data module is used to record data that is expected to be greater than a preset value as valid data; The data combining module is used to combine the valid data and the shared data to obtain the data clustering result corresponding to the data cluster group.

6. The data clustering integration and fusion device according to claim 5, characterized in that, The data cluster group module includes a first target coordinate module, a second target coordinate module, a target data group module, and a data cluster grouping module: The first target coordinate module is used to calculate the average value of the coordinates of each data in each data cluster to obtain the first target coordinates for each data cluster in each data cluster set; The second target coordinate module is used to average the centroid coordinates of the data cluster with the first target coordinates to obtain the second target coordinates; the centroid coordinates are generated by the clustering algorithm when classifying environmental monitoring data; The target data group module is used to acquire any data cluster set, and group the data clusters within the data cluster set to obtain target data groups; each target data group corresponds to a second target coordinate. The data cluster grouping module is used to determine the target distance between the second target coordinates corresponding to each data cluster in the data cluster set and the second target coordinates corresponding to the target data group for each remaining data cluster set. The data clusters in the data cluster set are grouped based on the minimum target distance value to obtain a data cluster group.

7. The data clustering integration and fusion device according to claim 5, characterized in that, The expected distribution module includes an invalid data module, a replacement data cluster module, a target expectation module, and a data classification module: The invalid data module is used to record data that is less than a preset value as invalid data; The replacement data cluster module is used to record the data cluster corresponding to the invalid data as the original data cluster, and for the original data cluster, to obtain the replacement data cluster that is closest to the original data cluster through the second target coordinates corresponding to the original data cluster; the replacement data cluster is the data cluster adjacent to the original data cluster. The target expectation module is used to classify the invalid data into a replacement data cluster and calculate the target expectation that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster. The data classification module is used to classify invalid data into the replacement data cluster if the target expectation is greater than the preset target expectation value.

8. The data clustering integration and fusion device according to claim 7, characterized in that, The target expectation module includes an invalid data partitioning module, a recalculation module, and a separate classification module: The invalid data partitioning module is used to partition the invalid data into the next nearest replacement data cluster if the target expectation is less than the preset target expectation value. The recalculation module is used to recalculate the target expectation that the invalid data falls on the multivariate normal distribution function corresponding to the replacement data cluster; The separate classification module is used to repeat the replacement data cluster module and the recalculation module as described above if the target expectation is still less than the preset target expectation value, until the target expectation of the invalid data is less than the preset target expectation value for all replacement data clusters, and then the invalid data is recorded as special data and classified separately.

Images