Data association method, device and system and storage medium

Abstract

The invention provides a data association method, device and system and a storage medium, and belongs to the technical field of data association, and the method comprises the steps: importing original geoscience space data, geoscience space coordinates, original non-space data and non-space coordinates; preprocessing the original geoscience spatial data and the original non-spatial data to obtain preprocessed geoscience spatial data and preprocessed non-spatial data; and screening out a target geoscience data set from the preprocessed geoscience space data and the preprocessed non-space data according to the geoscience space coordinates and the non-space coordinates. According to the method, the correlation efficiency of the geoscience spatial data and the non-spatial data is improved, the reliability and the accuracy of a correlation result are enhanced, and the requirement of data explosive growth can be met.

Description

Technical Field

[0001] This invention relates primarily to the field of data association technology, specifically to a data association method, apparatus, system, and storage medium. Background Technology

[0002] Geospatial data is a core element of Geographic Information Systems (GIS), including location information, topographic attributes, and soil characteristics. This data plays a crucial role in environmental monitoring (using satellite imagery and meteorological data to predict climate change), resource management, and disaster prediction (such as predicting landslides through climate and surface data). Non-spatial data typically refers to data without geographic location information, such as text, images, and time series data. This data is widely used in machine learning and data analysis. With the development of data science and technology, more and more application scenarios require the combined analysis of spatial and non-spatial data. Examples include the correlation between geographic features and non-geographic features (such as demographic data) in GIS, the use of satellite imagery (spatial data) and meteorological data (non-spatial data) for disaster prediction in environmental monitoring, and location-based recommendation systems that combine user location information (spatial data) with user behavior data (non-spatial data). Therefore, effectively extracting and utilizing the correlation between geospatial and non-spatial data has become a significant challenge.

[0003] In existing correlation techniques, the association between geospatial and non-spatial data typically relies on simple statistical analysis or rule-based pattern matching. This approach ignores the geographical characteristics and data lineage of geospatial data, resulting in inaccurate and scalable results. Furthermore, traditional correlation methods often only process data with single attributes and cannot comprehensively consider multi-dimensional spatial and non-spatial information, affecting the correlation effect. For data with spatiotemporal characteristics (such as video or time series), existing methods also struggle to capture dynamic temporal and spatial relationships. Faced with massive amounts of geospatial and non-spatial data, existing technologies suffer from inefficiency in computation, limiting their application scope.

[0004] Despite the rapid development of big data technology, the characteristics of spatial data (such as its distribution, heterogeneity, and complexity) pose many challenges to its correlation with other types of data. First, regarding data distribution and geographic specificity, spatial data typically exhibits a non-uniform distribution, with alternating high-density and low-density areas. This distribution can lead to biases in traditional statistical methods or machine learning models, especially in boundary or sparse areas. Furthermore, patterns in geospatial data often have strong geographic location dependencies (such as land cover types and climate characteristics), making it difficult for traditional data analysis methods to correlate geographic location information with other non-spatial data using geocoding techniques. Second, regarding data heterogeneity and complexity, spatial data analysis typically requires processing data from different sensors or platforms (such as satellite imagery and ground observation data). This data has different types (structured / unstructured), resolutions, and precisions, resulting in strong data heterogeneity. High-resolution spatial data often leads to the generation of high-dimensional feature spaces, and the data may be sparse in many dimensions. This "high-dimensional sparsity" problem can lead to inefficient model training and increase the risk of overfitting. Third, regarding data scale and spatiotemporal synchronization, spatial data analysis typically involves massive amounts of data (such as location data in Geographic Information Systems (GIS) and remote sensing images). Traditional methods may face problems of low computational efficiency and high memory consumption when processing large-scale data. Furthermore, large-scale spatial data analysis requires complex algorithmic support, such as the visualization and analysis of three-dimensional or four-dimensional spatial data, which all require significant computational resources. Many phenomena on the Earth's surface exhibit temporal and spatial synchronicity (such as surface changes and climate change), and may contain dynamically changing data. Therefore, how to balance the dynamic changes in both the temporal and spatial dimensions during analysis is a significant challenge in spatial data analysis. Fourth, regarding data privacy and security, spatial data may contain a large amount of location information or geographically sensitive data. When conducting spatial data analysis, it is essential not only to ensure the legality and technical compliance of data sharing but also to protect user privacy and data security. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a data association method, apparatus, system and storage medium to address the shortcomings of the prior art.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A data association method, comprising the following steps:

[0007] S1: Import multiple raw geospatial data, geospatial coordinates corresponding to each of the raw geospatial data, multiple raw non-spatial data, and non-spatial coordinates corresponding to each of the raw non-spatial data;

[0008] S2: Preprocess each of the original geospatial data and each of the original non-spatial data respectively to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data.

[0009] S3: Select multiple target geoscience data from all the preprocessed geoscience spatial data and all the preprocessed non-spatial data based on all the geoscience spatial coordinates and all the non-spatial coordinates, and combine all the target geoscience data to obtain the target geoscience dataset;

[0010] S4: Construct a deep learning model, and perform model analysis on the deep learning model using the target geoscience dataset to obtain a data association model;

[0011] S5: Import multiple geospatial data to be associated and multiple non-spatial data to be associated, and perform data association on all the geospatial data to be associated and all the non-spatial data to be associated through the data association model to obtain the data association result.

[0012] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: A data association device, comprising:

[0013] The import module is used to import multiple raw geospatial data, geospatial coordinates corresponding to each of the raw geospatial data, multiple raw non-spatial data, and non-spatial coordinates corresponding to each of the raw non-spatial data.

[0014] The preprocessing module is used to preprocess each of the original geospatial data and each of the original non-spatial data respectively to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data.

[0015] The filtering module is used to filter out multiple target geoscience data from all the preprocessed geoscience spatial data and all the preprocessed non-spatial data based on all the geoscience spatial coordinates and all the non-spatial coordinates, and to combine all the target geoscience data to obtain a target geoscience dataset;

[0016] The model analysis module is used to construct a deep learning model and perform model analysis on the deep learning model using the target geoscience dataset to obtain a data association model.

[0017] The import module is also used to import multiple geospatial data to be associated and multiple non-spatial data to be associated;

[0018] The result acquisition module is used to perform data association on all the geospatial data to be associated and all the non-spatial data to be associated through the data association model to obtain the data association result.

[0019] Based on the above-mentioned data association method, the present invention also provides a data association system.

[0020] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a data association system, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the data association method described above is implemented.

[0021] Based on the above-described data association method, the present invention also provides a computer-readable storage medium.

[0022] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the data association method as described above.

[0023] The beneficial effects of this invention are as follows: by preprocessing the original geospatial data and the original non-spatial data, preprocessed geospatial data and preprocessed non-spatial data are obtained; a target geospatial dataset is selected from the preprocessed geospatial data and the preprocessed non-spatial data based on geospatial coordinates and non-spatial coordinates; a data association model is obtained by analyzing the deep learning model using the target geospatial dataset; and a data association result is obtained by associating the geospatial data to be associated and the non-spatial data to be associated using the data association model. This improves the association efficiency between geospatial data and non-spatial data, enhances the reliability and accuracy of the association results, and can adapt to the needs of explosive data growth. Attached Figure Description

[0024] Figure 1 A flowchart illustrating the data association method provided in an embodiment of the present invention;

[0025] Figure 2 This is a schematic diagram of data lineage analysis for the data association method provided in this embodiment of the invention;

[0026] Figure 3 This is a schematic diagram of the data association model for the data association method provided in this embodiment of the invention;

[0027] Figure 4 A block diagram of a data association device provided in an embodiment of the present invention. Detailed Implementation

[0028] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0029] Figure 1 This is a flowchart illustrating a data association method provided in an embodiment of the present invention.

[0030] like Figure 1 As shown, a data association method includes the following steps:

[0031] S1: Import multiple raw geospatial data, geospatial coordinates corresponding to each of the raw geospatial data, multiple raw non-spatial data, and non-spatial coordinates corresponding to each of the raw non-spatial data;

[0032] S2: Preprocess each of the original geospatial data and each of the original non-spatial data respectively to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data.

[0033] S3: Select multiple target geoscience data from all the preprocessed geoscience spatial data and all the preprocessed non-spatial data based on all the geoscience spatial coordinates and all the non-spatial coordinates, and combine all the target geoscience data to obtain the target geoscience dataset;

[0034] S4: Construct a deep learning model, and perform model analysis on the deep learning model using the target geoscience dataset to obtain a data association model;

[0035] S5: Import multiple geospatial data to be associated and multiple non-spatial data to be associated, and perform data association on all the geospatial data to be associated and all the non-spatial data to be associated through the data association model to obtain the data association result.

[0036] It should be understood that this involves acquiring high-resolution geographic information system (GIS) data, geological exploration data and other geospatial data (i.e., raw geospatial data), and environmental monitoring data and other non-spatial data (i.e., raw non-spatial data).

[0037] In the above embodiments, preprocessing of the original geospatial data and original non-spatial data yields preprocessed geospatial data and preprocessed non-spatial data. Based on geospatial and non-spatial coordinates, a target geospatial dataset is selected from the preprocessed geospatial data and preprocessed non-spatial data. A data association model is obtained through model analysis of the deep learning model using the target geospatial dataset. The data association model is then used to associate the geospatial data to be associated with the non-spatial data to be associated with, resulting in a data association result. This improves the efficiency of associating geospatial and non-spatial data, enhances the reliability and accuracy of the association result, and can adapt to the needs of explosive data growth.

[0038] Optionally, as an embodiment of the present invention, the process of S2 includes:

[0039] Denoising is performed on each of the original geospatial data and each of the original non-spatial data to obtain denoised geospatial data corresponding to each of the original geospatial data and denoised non-spatial data corresponding to each of the original non-spatial data.

[0040] Outlier processing is performed on each of the denoised geospatial data and each of the denoised non-spatial data to obtain processed geospatial data corresponding to each of the original geospatial data and processed non-spatial data corresponding to each of the original non-spatial data.

[0041] Each of the processed geospatial data and each of the processed non-spatial data are projected onto a geographic coordinate system to obtain projected geospatial data corresponding to each of the original geospatial data and projected non-spatial data corresponding to each of the original non-spatial data.

[0042] Normalization is performed on each of the projected geospatial data and each of the projected non-spatial data to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data.

[0043] It should be understood that denoising is performed on geospatial data (i.e., raw geospatial data) and non-spatial data (i.e., raw non-spatial data) to remove outliers or missing values. Then, the different data sources (i.e., processed geospatial data and processed non-spatial data) are integrated into the same geographic coordinate system and normalized (e.g., standardization, classification) at the same spatial resolution.

[0044] In the above embodiments, preprocessing the original geospatial data and the original non-spatial data to obtain preprocessed geospatial data and preprocessed non-spatial data improves the correlation efficiency between geospatial data and non-spatial data and enhances the reliability and accuracy of the correlation results.

[0045] Optionally, as an embodiment of the present invention, in step S3, the process of selecting multiple target geoscientific data from all the preprocessed geoscientific spatial data and all the preprocessed non-spatial data based on all the geoscientific spatial coordinates and all the non-spatial coordinates includes:

[0046] Feature extraction is performed on each of the preprocessed geospatial data to obtain the first geospatial feature corresponding to each of the preprocessed geospatial data.

[0047] Attribute extraction is performed on each of the preprocessed non-spatial data to obtain the first non-spatial attribute corresponding to each of the preprocessed non-spatial data.

[0048] Based on mutual information, the first geospatial features corresponding to each of the preprocessed geospatial data, and the first non-spatial attributes corresponding to each of the preprocessed non-spatial data, multiple first-filtered geospatial data and multiple first-filtered non-spatial data are selected from all the preprocessed geospatial data and all the preprocessed non-spatial data.

[0049] Feature extraction is performed on each of the first-filtered geospatial data to obtain the second geospatial features corresponding to each of the first-filtered geospatial data.

[0050] Attribute extraction is performed on each of the first-filtered non-spatial data to obtain the second non-spatial attribute corresponding to each of the first-filtered non-spatial data.

[0051] Based on the Pearson correlation coefficient, the second geospatial features corresponding to each of the first-screened geospatial data, and the second non-spatial attributes corresponding to each of the first-screened non-spatial data, multiple second-screened geospatial data and multiple second-screened non-spatial data are selected from all the first-screened geospatial data and all the first-screened non-spatial data.

[0052] Feature extraction is performed on each of the second-filtered geospatial data to obtain the third geospatial features corresponding to each of the second-filtered geospatial data.

[0053] Attribute extraction is performed on each of the second-filtered non-spatial data to obtain the third non-spatial attribute corresponding to each of the second-filtered non-spatial data.

[0054] Based on cosine similarity, the third geospatial features corresponding to each of the second-selected geospatial data, and the third non-spatial attributes corresponding to each of the second-selected non-spatial data, multiple third-selected geospatial data and multiple third-selected non-spatial data are selected from all the second-selected geospatial data and all the second-selected non-spatial data.

[0055] Based on the geospatial coordinates corresponding to all the third-filtered geospatial data and the non-spatial coordinates corresponding to all the third-filtered non-spatial data, multiple fourth-filtered geospatial data and multiple fourth-filtered non-spatial data are selected from all the third-filtered geospatial data and all the third-filtered non-spatial data.

[0056] All the fourth-filtered geospatial data and all the fourth-filtered non-spatial data are used as target geospatial data, thus obtaining multiple target geospatial data.

[0057] It should be understood that attribute extraction refers to extracting attribute information from data.

[0058] Specifically, feature extraction is first performed on the geoscientific data (i.e., preprocessed geoscientific spatial data). Features related to geographical location (such as rock type, geological structure, etc.) (i.e., first geoscientific spatial features) are extracted from the geoscientific spatial data (i.e., preprocessed geoscientific spatial data), and attributes related to the target variable (i.e., first non-spatial attributes) are extracted from the non-spatial data (i.e., preprocessed non-spatial data). Then, mutual information, Pearson correlation coefficient, or cosine similarity are used to quantify the correlation between the two. By combining the use of these three methods, the degree of correlation between geoscientific spatial data and non-spatial data can be comprehensively evaluated, thereby measuring the degree of correlation between geoscientific spatial data and non-spatial data. In addition, it is also necessary to adjust the spatiotemporal weights of the data lineage measurement based on geospatial weights (distance decay, direction effect).

[0059] In the above embodiments, multiple target geoscience data are selected from all preprocessed geoscience spatial data and all preprocessed non-spatial data based on all geoscience spatial coordinates and all non-spatial coordinates. This improves the correlation efficiency between geoscience spatial data and non-spatial data, enhances the reliability and accuracy of the correlation results, can adapt to the needs of explosive data growth, and is applicable to various fields such as environmental monitoring, resource management, and disaster prediction.

[0060] Optionally, as an embodiment of the present invention, the first geospatial feature includes a plurality of first geospatial sub-features, and the first non-spatial attribute includes a plurality of first non-spatial sub-attributes.

[0061] The process of selecting multiple first-filtered geospatial data and multiple first-filtered non-spatial data from all the preprocessed geospatial data and all the preprocessed non-spatial data based on mutual information, the first geospatial features corresponding to each of the preprocessed geospatial data, and the first non-spatial attributes corresponding to each of the preprocessed non-spatial data includes:

[0062] By calculating the mutual information of multiple first geospatial sub-features in each of the first geospatial features and multiple first non-spatial sub-attributes in each of the first non-spatial attributes using the first formula, multiple mutual information corresponding to each of the first geospatial features is obtained. The first formula is:

[0063]

[0064] Among them, MI(X) a1 ;Y b1 ) represents the mutual information between the a1th first geospatial feature and the b1th first non-spatial attribute. For the a1th first geospatial feature, the i1th first geospatial sub-feature is... Let b1 be the j1st non-spatial sub-attribute of the first non-spatial attribute, and p() be the probability distribution function.

[0065] If the mutual information is greater than a preset first threshold, the preprocessed geospatial data corresponding to the mutual information is used as the first filtered geospatial data, and the first non-spatial sub-attribute corresponding to the mutual information is used as the first filtered non-spatial data, thereby obtaining multiple first filtered geospatial data and multiple first filtered non-spatial data.

[0066] Preferably, the preset first threshold can be 0.

[0067] Specifically, the mutual information formula is as follows:

[0068]

[0069] Where X represents geospatial data (such as rock type, geological structure, etc.) (i.e., the first geospatial feature), Y represents non-spatial data (i.e., the first non-spatial attribute), p(x,y) is the probability of X and Y occurring simultaneously (i.e., the joint probability distribution function), and p(x) and p(y) are the probabilities of X and Y occurring individually (i.e., the marginal probability distribution functions). Mutual information is used to measure the independence between geospatial data X (i.e., the first geospatial feature) and non-spatial data Y (i.e., the first non-spatial attribute). The larger the value, the stronger the correlation between the two; conversely, the weaker the correlation. In data lineage analysis, mutual information can be used to assess the degree of association between geospatial features and non-spatial variables at different geographical locations.

[0070] Specifically, the mutual information formula calculates the degree of correlation between the feature information extracted from spatial data (i.e., the first geospatial feature) and the attribute information extracted from non-spatial data (i.e., the first non-spatial attribute). When MI(X;Y)>0, it means that the two are correlated, and then the next calculation is performed. If the two are completely independent, then p(x,y)=p(x)p(y), and the division between the two is 1, so log1=0. Therefore, when the result is equal to 0, it means that the two are not correlated at all and are completely independent.

[0071] In the above embodiments, based on mutual information, first geospatial features, and first non-spatial attributes, first filtered geospatial data and first filtered non-spatial data are selected from preprocessed geospatial data and preprocessed non-spatial data. The correlation between geospatial features and non-spatial variables at different geographical locations is evaluated, which improves the correlation efficiency between geospatial data and non-spatial data and enhances the reliability and accuracy of the correlation results.

[0072] Optionally, as an embodiment of the present invention, the second geospatial feature includes a plurality of second geospatial sub-features, and the second non-spatial attribute includes a plurality of second non-spatial sub-attributes.

[0073] The process of selecting multiple second-selected geospatial data and multiple second-selected non-spatial data from all first-selected geospatial data and all first-selected non-spatial data based on Pearson correlation coefficient, second geospatial features corresponding to each of the first-selected geospatial data, and second non-spatial attributes corresponding to each of the first-selected non-spatial data includes:

[0074] By calculating the Pearson correlation coefficients for multiple second geospatial sub-features in each of the second geospatial features and multiple second non-spatial sub-attributes in each of the second non-spatial attributes using the second formula, multiple Pearson correlation coefficients corresponding to each of the second geospatial features are obtained. The second formula is:

[0075]

[0076] in, Let n be the Pearson correlation coefficient between the a2th second geospatial feature and the b2th second non-spatial attribute, and n1 be the total number of second geospatial sub-features. This is the i2th second geospatial sub-feature within the a2th second geospatial feature. This represents the average value of the a2nd second geoscientific spatial feature. For the b2nd second non-spatial attribute, the j2nd second non-spatial sub-attribute. b2 is the average value of the second non-spatial attribute, and m1 is the total number of the second non-spatial sub-attributes;

[0077] If the absolute value of the Pearson correlation coefficient is greater than a preset second threshold, then the first filtered geospatial data corresponding to the Pearson correlation coefficient is used as the second filtered geospatial data, and the first filtered non-spatial data corresponding to the Pearson correlation coefficient is used as the second filtered non-spatial data, thereby obtaining multiple second filtered geospatial data and multiple second filtered non-spatial data.

[0078] Preferably, the preset second threshold can be 0.3.

[0079] Specifically, the formula for the Pearson correlation coefficient is as follows:

[0080]

[0081] X represents geospatial data (i.e., second geospatial features), and Y represents non-spatial data (i.e., second non-spatial attributes). and These are the means of X and Y, respectively. The Pearson correlation coefficient measures the strength of a linear relationship, ranging from -1 to 1. Positive values ​​indicate a positive correlation, negative values ​​indicate a negative correlation, and the larger the absolute value, the stronger the correlation. In the association of geospatial and non-spatial data, the Pearson correlation coefficient can be used to assess the linear relationship between specific geospatial features and non-spatial variables at a given geographical location.

[0082] It should be understood that the Pearson coefficient is a statistical indicator that measures the linear correlation between two variables. When the absolute value of r, i.e., |r|... X,Y When |>0.3, it is included in the next calculation because |r X,Y |<0.3 indicates a weak linear correlation, 0.3<|r X,Y |<0.7 moderate correlation, |r X,Y A value of 0.7 indicates a strong correlation.

[0083] In the above embodiments, second-screened geospatial data and second-screened non-spatial data are selected from the first-screened geospatial data and the first-screened non-spatial data based on the Pearson correlation coefficient, the second geospatial feature, and the second non-spatial attribute. This evaluates the linear relationship between specific geospatial features and non-spatial variables at geographical locations and improves the correlation efficiency between geospatial data and non-spatial data.

[0084] Optionally, as an embodiment of the present invention, the third geospatial feature includes multiple third geospatial sub-features, and the third non-spatial attribute includes multiple third non-spatial sub-attributes.

[0085] The process of selecting multiple third-selected geospatial data and multiple third-selected non-spatial data from all second-selected geospatial data and all second-selected non-spatial data based on cosine similarity, the third geospatial features corresponding to each of the second-selected geospatial data, and the third non-spatial attributes corresponding to each of the second-selected non-spatial data includes:

[0086] The third equation is used to calculate the cosine similarity of multiple third geospatial sub-features in each of the aforementioned third geospatial features and multiple third non-spatial sub-attributes in each of the aforementioned third non-spatial attributes, thereby obtaining multiple cosine similarities corresponding to each of the aforementioned third geospatial features. The third equation is as follows:

[0087]

[0088] Among them, CosSim(X a3 ,Y b3 ) represents the Pearson correlation coefficient between the a3rd third geospatial feature and the b3rd third non-spatial attribute, and n2 represents the total number of third geospatial sub-features. This is the i3rd third geospatial sub-feature within the a3rd third geospatial feature. Let b be the j3rd third non-spatial sub-attribute in the b3rd third non-spatial attribute, and m2 be the total number of third non-spatial sub-attributes;

[0089] If the cosine similarity is greater than a preset third threshold, then the second-filtered geospatial data corresponding to the cosine similarity is used as the third-filtered geospatial data, and the second-filtered non-spatial data corresponding to the cosine similarity is used as the third-filtered non-spatial data, thereby obtaining multiple third-filtered geospatial data and multiple third-filtered non-spatial data.

[0090] Preferably, the preset third threshold can be 0.

[0091] Specifically, the cosine similarity formula is as follows:

[0092]

[0093] X and Y represent the vector forms of geospatial data (i.e., third geospatial features) and non-spatial data (i.e., third non-spatial attributes), respectively. X is... i and y i These are the values ​​of X and Y in the i-th dimension. Cosine similarity measures the angle between two vectors, with a value ranging from -1 to 1. A positive value indicates an angle less than 90°, and a negative value indicates an angle greater than 90°. In the association of geospatial and non-spatial data, cosine similarity can be used to evaluate the similarity between geographic feature vectors and non-spatial variable vectors.

[0094] It should be understood that the numerator is the inner product of the two vectors, which mainly reflects the "degree of coordination" between the two vectors in the same direction. The denominator is the product of the magnitudes of the two vectors. When the value = 1, it means that the two are completely similar; when the value = -1, it means that the two are completely opposite; and when the value = 0, it means that the two have no similarity. Therefore, taking CosSim(X,Y) > 0 into the next calculation indicates that the two have a certain degree of positive similarity.

[0095] In the above embodiments, third-selection geospatial data and third-selection non-spatial data are selected from the second-selection geospatial data and the second-selection non-spatial data based on cosine similarity, third geospatial features, and third non-spatial attributes. This evaluates the similarity between geospatial feature vectors and non-spatial variable vectors at geographical locations and improves the association efficiency between geospatial data and non-spatial data.

[0096] Optionally, as an embodiment of the present invention, the process of selecting a plurality of fourth-selected geospatial data and a plurality of fourth-selected non-spatial data from all the third-selected geospatial data and all the third-selected non-spatial data based on the geospatial coordinates corresponding to all the third-selected geospatial data and the non-spatial coordinates corresponding to all the third-selected non-spatial data includes:

[0097] Calculate the distance between the geospatial coordinates corresponding to each of the third-filtered geospatial data and the non-spatial coordinates corresponding to each of the third-filtered non-spatial data, and obtain multiple coordinate distances corresponding to each of the third-filtered geospatial data.

[0098] By using the fourth formula, the geospatial coordinates corresponding to each of the third-selected geospatial data, and the non-spatial coordinates corresponding to each of the third-selected non-spatial data, a comprehensive similarity calculation is performed on each of the third-selected geospatial data and each of the third-selected non-spatial data, respectively, to obtain multiple comprehensive similarities corresponding to each of the third-selected geospatial data. The fourth formula is:

[0099] in,

[0100] Among them, S a4b4 Let S0 be the overall similarity between the geospatial data after the a4 third screening and the non-spatial data after the b4 third screening, and let S0 be the cardinality similarity, ω() be the distance decay function, and M be the distance similarity between the two sets of data. a4b4 X is the method effect adjustment term for the geospatial data after the a4th third screening and the non-spatial data after the b4th third screening. a4 For the a4th geospatial data after the third screening, Y b4This is the fourth non-spatial data point after the third filtering. The rate of decay is exponential, where β0, β1, and β2 are all parameters of the first model, and d pq Let σ be the coordinate distance between the p-th geospatial coordinate and the q-th non-spatial coordinate, and let σ be the attenuation radius parameter.

[0101] If the overall similarity is greater than a preset fourth threshold, then the third-filtered geospatial data corresponding to the overall similarity is used as the fourth-filtered geospatial data, and the third-filtered non-spatial data corresponding to the overall similarity is used as the fourth-filtered non-spatial data, thereby obtaining multiple fourth-filtered geospatial data and multiple fourth-filtered non-spatial data.

[0102] Preferably, the preset fourth threshold can be 0.5.

[0103] Specifically, the distance attenuation formula is as follows:

[0104] S ij =S0×ω(d ij ),

[0105]

[0106] ω(d ij )=exp(-kd ij ),

[0107] In the above formula, S ij S0 represents the similarity or correlation between geographical locations i and j, where S0 is the cardinality similarity (when the distance d is less than or equal to the distance d). ij When = 0, i.e., similarity under the condition of the same geographical location), w(d ij w(d) is a distance decay function, indicating that the similarity gradually decreases as the geographical distance increases. ij This is used to quantify the decrease in similarity between geographical locations i and j due to spatial distance. Common distance decay functions include inverse distance decay, Gaussian decay, and exponential decay. Decay functions help reduce the impact of distant geographical locations on the correlation calculation, thereby improving the accuracy of the model. c is a smoothing parameter to prevent division by zero errors; k is the decay exponent, usually 2 or higher, representing the rate at which similarity decreases with increasing distance; σ is the decay radius parameter, controlling the smoothness of the decay; and k is the decay rate parameter, controlling the rate at which similarity decreases with increasing distance.

[0108] The formula for method effect is as follows:

[0109]

[0110] In the formula, M ijThe first term represents the method effect adjustment for geographic locations i and j. The second term is the method effect function, used to quantify the specific impact of a particular method on similarity calculation. ① represents the regional average effect, ② the time effect, and ③ the adjustment for exogenous variables. The method effect adjustment term introduces additional information or factors to help calculate the similarity between geographic locations i and j more accurately. This method can better reflect the complex spatial relationships in a specific application context. For example, in geostatistics, it may be necessary to consider the influence of regional effects or specific variables. The method effect function can be designed according to the specific application scenario.

[0111] Combining distance decay and method effects, the comprehensive similarity S between geographic locations i and j... i It can be expressed as the following formula:

[0112] S ij =S0×ω(d ij )+M ij ,

[0113] Adding the method effect adjustment term directly to the similarity after distance decay allows for a more comprehensive capture of the complex relationship between geographic locations i and j, including the impact of spatial distance and additional adjustment factors specific to the method. In the "data lineage" approach, the distance decay formula quantifies the reduction in similarity between geographic locations i and j due to spatial distance, while the method effect adjustment term introduces additional information or factors to help calculate similarity more accurately. The specific implementation can be tailored to different function forms (e.g., inverse distance decay, Gaussian decay) and adjustment terms (e.g., regional average effect, time effect) depending on the application requirements. Combining these formulas allows for better modeling and analysis of the relationship between geospatial and non-spatial data.

[0114] It should be understood that while the method effect can calculate the similarity between two different geographical locations, it is similar to neural network learning. It requires adding attention mechanisms to adjust the weights of important attributes or regions to avoid missing crucial information. Therefore, adding the method effect to distance decay can better capture the complex relationship between the two. Summations or products can be calculated; when the sum is greater than 0.5, it is included in the next calculation.

[0115] In the above embodiments, geospatial data and non-spatial data are filtered from the third filtered geospatial data and the third filtered non-spatial data based on geospatial coordinates and non-spatial coordinates, so as to avoid missing important information and to calculate the similarity between geographical locations more accurately.

[0116] Optionally, as an embodiment of the present invention, in step S4, the process of performing model analysis on the deep learning model using the target geoscience dataset to obtain a data association model includes:

[0117] S41: Initialize the regularization parameters and the second model parameters;

[0118] S42: Use a genetic algorithm to optimize the initialized regularization parameters to obtain the optimized regularization parameters;

[0119] S43: Perform feature extraction on the geoscience data in the target geoscience dataset respectively, and obtain the feature matrix through the result of feature extraction;

[0120] S44: Optimize the feature matrix, the optimized regularization parameters, and the initialized second model parameters using the fifth equation to obtain the optimized second model parameters. The fifth equation is:

[0121]

[0122] in, Here are the optimized second model parameters: y is the preset target variable, X′ is the feature matrix, ||||1 is the first norm, θ is the initialized second model parameters, λ is the optimized regularization parameter, and ||||2 is the second norm.

[0123] S45: Determine whether the optimized second model parameters are less than or equal to the preset model parameter threshold. If yes, execute S46; otherwise, return to S1.

[0124] S46: Train the target geoscience dataset using a deep learning model to obtain multiple predicted values;

[0125] S47: Return to S1 until the iteration number is reached, thereby obtaining multiple predicted values ​​corresponding to the iteration number;

[0126] S48: Import multiple true values ​​corresponding to the number of iterations, and calculate the mean of the loss function using the sixth equation on all predicted values ​​and all true values. The sixth equation is:

[0127]

[0128] in,

[0129] Where CV is the mean of the loss function, and MSE is... u Let z be the mean square error corresponding to the u-th iteration, k be the iteration number, w be the total number of predicted values, and z be the mean square error. o For the o-th true value, This is the o-th predicted value;

[0130] S49: Determine whether the mean of the loss function is less than or equal to the preset loss function threshold. If yes, execute S410; otherwise, return to S1.

[0131] S410: Update the parameters of the deep learning model based on the mean of the loss function to obtain a data association model.

[0132] Understandably, the process begins with feature selection of the geoscientific data. This involves using a feature selection method based on data lineage metrics, combining genetic algorithms with LASSO regression to screen for the optimal combination of spatial and non-spatial geoscientific features. This ensures that the selected features exhibit high relevance and stable spatial distribution characteristics. Next, the model is trained using machine learning or deep learning models (such as support vector machines, random forests, and convolutional neural networks) to model the association between spatial and non-spatial geoscientific data (i.e., the target geoscientific dataset). Input features include spatial geoscientific features and spatiotemporal weighted fusion features obtained through data lineage analysis. Finally, the model is optimized using cross-validation to select the optimal model parameters (such as regularization coefficients and tree depth).

[0133] Specifically, the basic formula for LASSO regression is:

[0134] The genetic algorithm optimizes λ: f(λ) = MSE(λ).

[0135] In the LASSO formula, λ is the regularization parameter, and the genetic algorithm is used to globally search for the optimal λ to minimize the prediction error. y is the target variable, X is the feature matrix, θ is the model parameter, and ||θ||1 is the first norm of θ, which is the sum of the absolute values ​​of all elements.

[0136] It should be understood that in machine learning, LASSO regression is a commonly used regularization method that limits the complexity of the model by adding a regularization term to the loss function. Its core idea is to optimize the objective function to make the model's parameters as sparse as possible, thereby preserving key features and simplifying the model.

[0137] Specifically, the predictive performance of the model can be determined by analyzing its prediction results. Therefore, the mean squared error (MSE) (Equation 5.9), cross-validation (Equation 5.8), and coefficient of determination (R²) are used. 2 The performance of a model is evaluated using metrics such as mean absolute error (MAE).

[0138]

[0139] Where n is the size of the dataset, yi It is the actual value. This is the predicted value. k is the number of folds, MSE. i This is the mean squared error (MSE) of the i-th fold. MSE is a commonly used loss function in regression problems, measuring the average squared difference between predicted and actual values. A smaller MSE value indicates more accurate model predictions. Due to the use of squared terms, MSE is highly sensitive to outliers (extreme values). Cross-validation is a method for evaluating model performance. It involves dividing the dataset into k subsets (folds), testing on each subset, and training on the remaining k-1 subsets. The final model performance is the average of all fold results. Cross-validation helps estimate the model's performance on independent data and reduces the risk of overfitting.

[0140] It should be understood that cross-validation (here, K-fold cross-validation) is used to address the high sensitivity of the squared term of the mean squared error. K-fold cross-validation reduces variance by averaging the results of training with k different groups, thus making the model's performance less sensitive to data partitioning. During training and testing, the model's training parameters are continuously optimized by minimizing the cross-validation (CV) value to achieve better training results; therefore, the CV does not have a fixed value.

[0141] In the above embodiments, the data association model is obtained by performing model analysis on the deep learning model using the target geoscience dataset, which reduces the risk of overfitting, makes the model parameters as sparse as possible, preserves key features, and simplifies the model.

[0142] Optionally, as another embodiment of the present invention, data lineage is the relationship formed between data throughout the entire data lifecycle, similar to human kinship. Data lineage is hierarchical, allowing for the tracing of data sources and destinations, which greatly aids in data quality improvement, data value assessment, and data lifecycle management. Data lineage analysis can be applied to the lineage analysis of datasets, helping to understand the lineage relationships between data fields, tables, or databases and other datasets. Furthermore, data lineage analysis methods include comprehensiveness, static analysis, contact-based analysis, logical temporal analysis, and timeliness. These methods help ensure the consistency of data flow tracking and achieve an objective reflection of data circulation.

[0143] This invention aims to establish a data lineage network by using a data lineage-based technology to associate geospatial data with non-spatial data through their inherent spatiotemporal distribution characteristics and attribute information. Data lineage includes multi-dimensional features such as geographical distribution, time series, and attribute similarity. Data lineage can effectively extract deep correlations between geospatial data and non-spatial data. The specific technical solution is as follows: (1) Data preprocessing stage. This mainly includes data cleaning and standardization. Geospatial data and non-spatial data are respectively subjected to preprocessing work such as denoising, filling missing values ​​(through linear interpolation, nearest neighbor interpolation), and normalization (scaling the data to the range of 0-1 to make features at different scales comparable) using filters or statistical methods to ensure data quality. For geospatial data, key features such as rock type, geological structure, and stratum thickness are extracted. These features are usually related to geographical location and have a potential impact on non-spatial variables; for non-spatial data, identify variables related to geographical location; and geocode the geographical location information in non-spatial data so that it can be associated with geospatial data. (2) Identification and extraction of geospatial features. This includes extracting representative features from geospatial data, such as rock type, stratigraphic thickness, geological structure, etc., and identifying the impact of these features on non-spatial data variables through machine learning models, as well as identifying variables that are significantly related to the target variable in non-spatial data through statistical analysis or machine learning methods (such as regression analysis, principal component analysis or cluster analysis). (3) Data association mechanism based on data lineage between geospatial data and non-spatial data. By calculating the correlation between geospatial features and non-spatial variables, it is determined whether there is a significant association between them. This mainly involves using data mining techniques to calculate the degree of association between geospatial features and non-spatial variables, and using similarity or distance measurement methods (such as cosine similarity, Pearson correlation coefficient, Euclidean distance, etc.) to calculate the degree of association between geospatial features and non-spatial variables. For high-dimensional data, dimensionality reduction techniques (such as principal component analysis) may be used to reduce dimensionality and simplify calculations. Then, statistical hypothesis testing methods (such as t-test, chi-square test, or F-test) are used to verify the significant association between geospatial features and non-spatial variables. By calculating the p-value, it is determined whether there is a statistically significant relationship between the feature and the variable. For example, p < 0.05 indicates a significant association. In addition, in spatial data, the influence of geographical location on variables may be non-uniform. Therefore, methods such as geographically weighted regression (GWR) can be used to calculate the local correlation of geospatial features at different locations, which can reveal the spatial heterogeneity between geospatial features and non-spatial variables. (4) Model training and mapping.By constructing mathematical models, a quantitative relationship is established between geoscientific spatial features and non-spatial variables. This step is the core task for achieving data association. First, regression analysis is performed. In simple cases, linear regression models are used to directly model the relationship between geoscientific features and non-spatial variables as a linear form. In complex cases, multinomial regression or locally weighted regression (LWR) may be needed to capture non-linear relationships. Second, machine learning models are built, using models such as Support Vector Regression (SVR), Random Forest Regression, or neural networks to non-linearly model the relationship between geoscientific features and non-spatial variables. These methods can handle complex relationships between features and demonstrate better predictive performance on large datasets. Finally, the model is optimized. During training, cross-validation (e.g., k-fold cross-validation) is used to select optimal hyperparameters (such as regularization coefficients, tree depth, etc.) to avoid overfitting. Then, mean squared error (MSE) and coefficient of determination (R²) are used. 2 Model performance can be evaluated using metrics such as mean absolute error (MAE), demonstrating the significant relationship between geospatial features and non-spatial variables. Data association mechanisms, through systematic preprocessing, feature extraction, and model training, establish quantitative relationships between geospatial features and non-spatial variables. This method effectively reveals the impact of geographical location on non-spatial variables and provides support for research and practice in related fields.

[0144] Alternatively, as another embodiment of the present invention, the beneficial effects of the present invention are as follows:

[0145] (1) High efficiency: Through data lineage analysis and deep learning algorithms, the correlation efficiency between geospatial data and non-spatial data has been significantly improved;

[0146] (2) Accuracy: The reliability and accuracy of the results are enhanced by using multi-dimensional feature extraction and association rule mining;

[0147] (3) Wide applicability: Applicable to a variety of fields, such as environmental monitoring, resource management, disaster prediction, etc.

[0148] (4) High scalability: It can handle massive and complex datasets and adapt to the needs of future data explosion.

[0149] Optionally, as another embodiment of the present invention, geospatial data typically includes geographically related features (such as rock type, geological structure, landform, etc.), while non-spatial data may be attributes related to these geographical locations (such as various environmental monitoring data). Traditional correlation techniques often struggle to utilize the spatiotemporal characteristics of both types of data simultaneously. Therefore, a method based on "data lineage" is proposed, aiming to achieve more effective correlation and prediction by mining the intrinsic connections between geospatial and non-spatial data.

[0150] Alternatively, as another embodiment of the present invention, such as Figure 2 As shown, the steps of the data lineage analysis in this invention are as follows:

[0151] First, feature extraction is performed on the geoscientific data. Features related to geographical location (such as rock type and geological structure) are extracted from the geoscientific spatial data, and attributes related to the target variable are extracted from the non-spatial data. Then, mutual information, Pearson correlation coefficient, or cosine similarity are used to quantify the correlation between the two. By combining the three methods, the degree of correlation between geoscientific spatial data and non-spatial data can be comprehensively evaluated, thereby measuring the degree of correlation between geoscientific spatial data and non-spatial data. In addition, it is necessary to adjust the spatiotemporal weights of the data lineage measurement based on geospatial weights (distance decay, direction effect).

[0152] Alternatively, as another embodiment of the present invention, the data migration and fusion steps of the present invention are as follows:

[0153] Based on data lineage measurement, geospatial and non-spatial data are migrated or merged geographically. In the target area, geospatial data is used to fill gaps in non-spatial data, or non-spatial data is used to correct geospatial data. For example, when processing soil and meteorological data, the similarity weight between each target geographic location and all reference geographic locations can be calculated first. Then, a weighted average of relevant data (such as meteorological parameters and soil characteristics) is performed based on the weights. The merged target geographic location data is then combined with other relevant data (such as historical yield records) to train a prediction model.

[0154] Alternatively, as another embodiment of the present invention, such as Figure 3 As shown, the construction and optimization steps of the model of this invention are as follows:

[0155] First, geoscientific data feature selection is performed using a feature selection method based on data lineage measurement, combining genetic algorithms with LASSO regression to screen for the optimal combination of spatial and non-spatial geoscientific features, ensuring that the selected features have high correlation and stable spatial distribution characteristics. Second, model training is conducted using machine learning or deep learning models (such as support vector machines, random forests, convolutional neural networks, etc.) to model the association between spatial and non-spatial geoscientific data. Input features include spatial geoscientific features and spatiotemporal weighted fusion features obtained through data lineage analysis. Finally, the model needs to be optimized using cross-validation to select the optimal model parameters (such as regularization coefficients, tree depth, etc.).

[0156] Figure 4This is a block diagram of a data association device provided in an embodiment of the present invention.

[0157] Alternatively, as another embodiment of the present invention, such as Figure 4 As shown, a data association device includes:

[0158] The import module is used to import multiple raw geospatial data, geospatial coordinates corresponding to each of the raw geospatial data, multiple raw non-spatial data, and non-spatial coordinates corresponding to each of the raw non-spatial data.

[0159] The preprocessing module is used to preprocess each of the original geospatial data and each of the original non-spatial data respectively to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data.

[0160] The filtering module is used to filter out multiple target geoscience data from all the preprocessed geoscience spatial data and all the preprocessed non-spatial data based on all the geoscience spatial coordinates and all the non-spatial coordinates, and to combine all the target geoscience data to obtain a target geoscience dataset;

[0161] The model analysis module is used to construct a deep learning model and perform model analysis on the deep learning model using the target geoscience dataset to obtain a data association model.

[0162] The import module is also used to import multiple geospatial data to be associated and multiple non-spatial data to be associated;

[0163] The result acquisition module is used to perform data association on all the geospatial data to be associated and all the non-spatial data to be associated through the data association model to obtain the data association result.

[0164] Optionally, another embodiment of the present invention provides a data association system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the data association method as described above. This system can be a computer or similar system.

[0165] Optionally, another embodiment of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the data association method as described above.

[0166] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0167] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0168] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0169] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0170] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0171] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0172] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A data association method, characterized in that, Includes the following steps: S1: Import multiple raw geospatial data, geospatial coordinates corresponding to each of the raw geospatial data, multiple raw non-spatial data, and non-spatial coordinates corresponding to each of the raw non-spatial data; S2: Preprocess each of the original geospatial data and each of the original non-spatial data respectively to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data. S3: Select multiple target geoscience data from all the preprocessed geoscience spatial data and all the preprocessed non-spatial data based on all the geoscience spatial coordinates and all the non-spatial coordinates, and combine all the target geoscience data to obtain the target geoscience dataset; S4: Construct a deep learning model, and perform model analysis on the deep learning model using the target geoscience dataset to obtain a data association model; S5: Import multiple geospatial data to be associated and multiple non-spatial data to be associated, and perform data association on all the geospatial data to be associated and all the non-spatial data to be associated through the data association model to obtain the data association result.

2. The data association method according to claim 1, characterized in that, The process of S2 includes: Denoising is performed on each of the original geospatial data and each of the original non-spatial data to obtain denoised geospatial data corresponding to each of the original geospatial data and denoised non-spatial data corresponding to each of the original non-spatial data. Outlier processing is performed on each of the denoised geospatial data and each of the denoised non-spatial data to obtain processed geospatial data corresponding to each of the original geospatial data and processed non-spatial data corresponding to each of the original non-spatial data. Each of the processed geospatial data and each of the processed non-spatial data are projected onto a geographic coordinate system to obtain projected geospatial data corresponding to each of the original geospatial data and projected non-spatial data corresponding to each of the original non-spatial data. Normalization is performed on each of the projected geospatial data and each of the projected non-spatial data to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data.

3. The data association method according to claim 1, characterized in that, In step S3, the process of selecting multiple target geoscientific data from all the preprocessed geoscientific spatial data and all the preprocessed non-spatial data based on all the geoscientific spatial coordinates and all the non-spatial coordinates includes: Feature extraction is performed on each of the preprocessed geospatial data to obtain the first geospatial feature corresponding to each of the preprocessed geospatial data. Attribute extraction is performed on each of the preprocessed non-spatial data to obtain the first non-spatial attribute corresponding to each of the preprocessed non-spatial data. Based on mutual information, the first geospatial features corresponding to each of the preprocessed geospatial data, and the first non-spatial attributes corresponding to each of the preprocessed non-spatial data, multiple first-filtered geospatial data and multiple first-filtered non-spatial data are selected from all the preprocessed geospatial data and all the preprocessed non-spatial data. Feature extraction is performed on each of the first-filtered geospatial data to obtain the second geospatial features corresponding to each of the first-filtered geospatial data. Attribute extraction is performed on each of the first-filtered non-spatial data to obtain the second non-spatial attribute corresponding to each of the first-filtered non-spatial data. Based on the Pearson correlation coefficient, the second geospatial features corresponding to each of the first-screened geospatial data, and the second non-spatial attributes corresponding to each of the first-screened non-spatial data, multiple second-screened geospatial data and multiple second-screened non-spatial data are selected from all the first-screened geospatial data and all the first-screened non-spatial data. Feature extraction is performed on each of the second-filtered geospatial data to obtain the third geospatial features corresponding to each of the second-filtered geospatial data. Attribute extraction is performed on each of the second-filtered non-spatial data to obtain the third non-spatial attribute corresponding to each of the second-filtered non-spatial data. Based on cosine similarity, the third geospatial features corresponding to each of the second-selected geospatial data, and the third non-spatial attributes corresponding to each of the second-selected non-spatial data, multiple third-selected geospatial data and multiple third-selected non-spatial data are selected from all the second-selected geospatial data and all the second-selected non-spatial data. Based on the geospatial coordinates corresponding to all the third-filtered geospatial data and the non-spatial coordinates corresponding to all the third-filtered non-spatial data, multiple fourth-filtered geospatial data and multiple fourth-filtered non-spatial data are selected from all the third-filtered geospatial data and all the third-filtered non-spatial data. All the fourth-filtered geospatial data and all the fourth-filtered non-spatial data are used as target geospatial data, thus obtaining multiple target geospatial data.

4. The data association method according to claim 3, characterized in that, The first geospatial feature includes multiple first geospatial sub-features, and the first non-spatial attribute includes multiple first non-spatial sub-attributes. The process of selecting multiple first-filtered geospatial data and multiple first-filtered non-spatial data from all the preprocessed geospatial data and all the preprocessed non-spatial data based on mutual information, the first geospatial features corresponding to each of the preprocessed geospatial data, and the first non-spatial attributes corresponding to each of the preprocessed non-spatial data includes: By calculating the mutual information of multiple first geospatial sub-features in each of the first geospatial features and multiple first non-spatial sub-attributes in each of the first non-spatial attributes using the first formula, multiple mutual information corresponding to each of the first geospatial features is obtained. The first formula is: Among them, MI(X) a1 ;Y b1 ) represents the mutual information between the a1th first geospatial feature and the b1th first non-spatial attribute. For the a1th first geospatial feature, the i1th first geospatial sub-feature is... Let b1 be the j1st non-spatial sub-attribute of the first non-spatial attribute, and p() be the probability distribution function. If the mutual information is greater than a preset first threshold, the preprocessed geospatial data corresponding to the mutual information is used as the first filtered geospatial data, and the first non-spatial sub-attribute corresponding to the mutual information is used as the first filtered non-spatial data, thereby obtaining multiple first filtered geospatial data and multiple first filtered non-spatial data.

5. The data association method according to claim 3, characterized in that, The second geospatial feature includes multiple second geospatial sub-features, and the second non-spatial attribute includes multiple second non-spatial sub-attributes. The process of selecting multiple second-selected geospatial data and multiple second-selected non-spatial data from all first-selected geospatial data and all first-selected non-spatial data based on Pearson correlation coefficient, second geospatial features corresponding to each of the first-selected geospatial data, and second non-spatial attributes corresponding to each of the first-selected non-spatial data includes: By calculating the Pearson correlation coefficients for multiple second geospatial sub-features in each of the second geospatial features and multiple second non-spatial sub-attributes in each of the second non-spatial attributes using the second formula, multiple Pearson correlation coefficients corresponding to each of the second geospatial features are obtained. The second formula is: in, Let a2 be the Pearson correlation coefficient between the second geospatial feature and the second non-spatial attribute, and n1 be the total number of second geospatial sub-features. This is the i2th second geospatial sub-feature within the a2th second geospatial feature. This represents the average value of the a2nd second geoscientific spatial feature. For the b2nd second non-spatial attribute, the j2nd second non-spatial sub-attribute. b2 is the average value of the second non-spatial attribute, and m1 is the total number of the second non-spatial sub-attributes; If the absolute value of the Pearson correlation coefficient is greater than a preset second threshold, then the first filtered geospatial data corresponding to the Pearson correlation coefficient is used as the second filtered geospatial data, and the first filtered non-spatial data corresponding to the Pearson correlation coefficient is used as the second filtered non-spatial data, thereby obtaining multiple second filtered geospatial data and multiple second filtered non-spatial data.

6. The data association method according to claim 3, characterized in that, The third geospatial feature includes multiple third geospatial sub-features, and the third non-spatial attribute includes multiple third non-spatial sub-attributes. The process of selecting multiple third-selected geospatial data and multiple third-selected non-spatial data from all second-selected geospatial data and all second-selected non-spatial data based on cosine similarity, the third geospatial features corresponding to each of the second-selected geospatial data, and the third non-spatial attributes corresponding to each of the second-selected non-spatial data includes: The third equation is used to calculate the cosine similarity of multiple third geospatial sub-features in each of the aforementioned third geospatial features and multiple third non-spatial sub-attributes in each of the aforementioned third non-spatial attributes, thereby obtaining multiple cosine similarities corresponding to each of the aforementioned third geospatial features. The third equation is as follows: Among them, CosSim(X a3 ,Y b3 ) represents the Pearson correlation coefficient between the a3rd third geospatial feature and the b3rd third non-spatial attribute, and n2 represents the total number of third geospatial sub-features. This is the i3rd third geospatial sub-feature within the a3rd third geospatial feature. Let b be the j3rd third non-spatial sub-attribute in the b3rd third non-spatial attribute, and m2 be the total number of third non-spatial sub-attributes; If the cosine similarity is greater than a preset third threshold, then the second-filtered geospatial data corresponding to the cosine similarity is used as the third-filtered geospatial data, and the second-filtered non-spatial data corresponding to the cosine similarity is used as the third-filtered non-spatial data, thereby obtaining multiple third-filtered geospatial data and multiple third-filtered non-spatial data.

7. The data association method according to claim 3, characterized in that, The process of selecting multiple fourth-selection geospatial data and multiple fourth-selection non-spatial data from all the third-selection geospatial data and all the third-selection non-spatial data based on the geospatial coordinates corresponding to all the third-selection geospatial data and the non-spatial coordinates corresponding to all the third-selection non-spatial data includes: Calculate the distance between the geospatial coordinates corresponding to each of the third-filtered geospatial data and the non-spatial coordinates corresponding to each of the third-filtered non-spatial data, and obtain multiple coordinate distances corresponding to each of the third-filtered geospatial data. By using the fourth formula, the geospatial coordinates corresponding to each of the third-selected geospatial data, and the non-spatial coordinates corresponding to each of the third-selected non-spatial data, a comprehensive similarity calculation is performed on each of the third-selected geospatial data and each of the third-selected non-spatial data, respectively, to obtain multiple comprehensive similarities corresponding to each of the third-selected geospatial data. The fourth formula is: S a4b4 =S0×ω(d pq )+M a4b4 , in, Among them, S a4b4 Let S0 be the overall similarity between the geospatial data after the a4 third screening and the non-spatial data after the b4 third screening, and let S0 be the cardinality similarity, ω() be the distance decay function, and M be the distance similarity between the two sets of data. a4b4 X is the method effect adjustment term for the geospatial data after the a4th third screening and the non-spatial data after the b4th third screening. a4 For the a4th geospatial data after the third screening, Y b4 This is the fourth non-spatial data point after the third filtering. The rate of decay is exponential, where β0, β1, and β2 are all parameters of the first model, and d pq Let σ be the coordinate distance between the p-th geospatial coordinate and the q-th non-spatial coordinate, and let σ be the attenuation radius parameter. If the overall similarity is greater than a preset fourth threshold, then the third-filtered geospatial data corresponding to the overall similarity is used as the fourth-filtered geospatial data, and the third-filtered non-spatial data corresponding to the overall similarity is used as the fourth-filtered non-spatial data, thereby obtaining multiple fourth-filtered geospatial data and multiple fourth-filtered non-spatial data.

8. The data association method according to claim 1, characterized in that, In step S4, the process of performing model analysis on the deep learning model using the target geoscience dataset to obtain a data association model includes: S41: Initialize the regularization parameters and the second model parameters; S42: Use a genetic algorithm to optimize the initialized regularization parameters to obtain the optimized regularization parameters; S43: Perform feature extraction on the geoscience data in the target geoscience dataset respectively, and obtain the feature matrix through the result of feature extraction; S44: Optimize the feature matrix, the optimized regularization parameters, and the initialized second model parameters using the fifth equation to obtain the optimized second model parameters. The fifth equation is: in, Here are the optimized second model parameters: y is the preset target variable, X′ is the feature matrix, ||||1 is the first norm, θ is the initialized second model parameters, λ is the optimized regularization parameter, and ||||2 is the second norm. S45: Determine whether the optimized second model parameters are less than or equal to the preset model parameter threshold. If yes, execute S46; otherwise, return to S1. S46: Train the target geoscience dataset using a deep learning model to obtain multiple predicted values; S47: Return to S1 until the iteration number is reached, thereby obtaining multiple predicted values ​​corresponding to the iteration number; S48: Import multiple true values ​​corresponding to the number of iterations, and calculate the mean of the loss function using the sixth equation on all predicted values ​​and all true values. The sixth equation is: in, Where CV is the mean of the loss function, and MSE is... u Let z be the mean square error corresponding to the u-th iteration, k be the iteration number, w be the total number of predicted values, and z be the mean square error. o For the o-th true value, This is the o-th predicted value; S49: Determine whether the mean of the loss function is less than or equal to the preset loss function threshold. If yes, execute S410; otherwise, return to S1. S410: Update the parameters of the deep learning model based on the mean of the loss function to obtain a data association model.

9. A data association device, characterized in that, include: The import module is used to import multiple raw geospatial data, geospatial coordinates corresponding to each of the raw geospatial data, multiple raw non-spatial data, and non-spatial coordinates corresponding to each of the raw non-spatial data. The preprocessing module is used to preprocess each of the original geospatial data and each of the original non-spatial data respectively to obtain preprocessed geospatial data corresponding to each of the original geospatial data and preprocessed non-spatial data corresponding to each of the original non-spatial data. The filtering module is used to filter out multiple target geoscience data from all the preprocessed geoscience spatial data and all the preprocessed non-spatial data based on all the geoscience spatial coordinates and all the non-spatial coordinates, and to combine all the target geoscience data to obtain a target geoscience dataset; The model analysis module is used to construct a deep learning model and perform model analysis on the deep learning model using the target geoscience dataset to obtain a data association model. The import module is also used to import multiple geospatial data to be associated and multiple non-spatial data to be associated; The result acquisition module is used to perform data association on all the geospatial data to be associated and all the non-spatial data to be associated through the data association model to obtain the data association result.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the data association method as described in any one of claims 1 to 8.

Images