Multi-source data aggregation method and system for realizing hierarchical query

Abstract

The invention discloses a multi-source data aggregation method and system for realizing hierarchical query, and relates to the technical field of data processing. The method comprises the steps of obtaining multi-source data of each storage unit of a target enterprise, and performing characterization according to the multi-source data to obtain multi-source features; substituting the multi-source features into an intelligent function model to obtain feature function scores; performing data important hierarchy division on the multi-source features through feature function scores to obtain a plurality of hierarchies; carrying out aggregation optimization on the multi-source features of any same level of each storage unit to obtain the aggregation credibility of each storage unit; performing abnormal feature filtering on the multi-source feature according to the aggregation credibility of each storage unit to obtain a first multi-source feature; performing weighted aggregation calculation according to the aggregation credibility to obtain an aggregation weight, and performing weighted averaging calculation according to the aggregation weight and the first multi-source feature to obtain an aggregation feature; generating an aggregation result according to the aggregation characteristics; according to the invention, the aggregation efficiency and accuracy of the multi-source data are improved.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, specifically relating to a method and system for hierarchical querying of multi-source data aggregation. Background Technology

[0002] In modern information systems, data processing is responsible for integrating and correlating heterogeneous data from different sources and formats, ensuring that this data can be transformed into unified, high-value information resources. This technology, through data cleaning, feature extraction, spatiotemporal alignment, and weighted fusion, transforms scattered multi-source data into datasets with consistency and correlation, and constructs multi-layered fusion models to support upper-level applications, providing reliable data support for intelligent decision-making in enterprises, efficient operation and maintenance in industries, and refined management in cities.

[0003] Existing technology (CN107066534B) discloses a multi-source data aggregation method and system, which involves: acquiring original data sources; obtaining an original strategy list based on the original data sources; sorting the original strategies in the original strategy list to form an inter-data source strategy list; aggregating the related domains in the original strategy list to generate an intra-data source aggregation list; sequentially acquiring each original strategy in the inter-data source strategy list; determining the left-side target data source and the right-side target data source corresponding to the original strategy based on the intra-data source aggregation list; and aggregating and outputting the data corresponding to the left-side target data source and the data corresponding to the right-side target data source. This method combines transaction flows from multiple databases within a relatively short time by searching and aggregating data from different databases using a prescribed strategy.

[0004] However, in the aforementioned patent, data from different databases is aggregated according to a prescribed strategy. The database corresponds to a storage unit. However, in practical applications, the layering of multi-source data within the storage unit still has defects, resulting in inaccurate aggregation of multi-source data and thus reducing the aggregation efficiency of multi-source data. Summary of the Invention

[0005] The purpose of this invention is to address the problem that the hierarchical structure of multi-source data within a storage unit is flawed in practical applications, which makes it impossible to aggregate abnormal data and thus reduces the efficiency of multi-source data aggregation. Therefore, this invention proposes a method and system for hierarchical querying of multi-source data aggregation.

[0006] In a first aspect of this invention, a method for multi-source data aggregation to achieve hierarchical querying is first proposed, the method comprising: Acquire multi-source data from each storage unit of the target enterprise, and perform feature extraction based on the multi-source data to obtain multi-source features; The feature function score is obtained by substituting the multi-source features into the intelligent function model; The data importance hierarchy of multi-source features is divided into multiple levels by feature function scoring; the aggregation and optimization of multi-source features of any level in each storage unit is performed to obtain the aggregation credibility of each storage unit. The first multi-source feature is obtained by filtering out abnormal features based on the aggregation credibility of each storage unit. The aggregation weight is obtained by weighted aggregation calculation based on the aggregation credibility, and the aggregation feature is obtained by weighted averaging of the aggregation weight and the first multi-source feature. The aggregation result is generated based on the aggregation characteristics; Obtain the aggregated results of each storage unit across all levels.

[0007] Optionally, the training process of the intelligent functional model includes: Multi-source features and feature function scores are obtained from the database. The multi-source features include sales features, cost features, and profit features. The feature function scores are obtained by experts through integrated functional evaluation based on the multi-source features. The intelligent functional model consists of functional layers of standard convolutional layers, convolutional-fully connected transition layers, and fully connected layers. The multi-source features and their corresponding feature function scores are integrated into several training and testing datasets. Several training datasets are imported into the artificial intelligence model for training. Test data is then used to test the trained AI model. Specifically, the multi-source features from the test data are input into the trained AI model, which outputs a feature function score. The absolute value of the difference between the feature function score and the feature function score recorded in the test data is checked to see if it is within an acceptable range. If yes, the test data set passes the test, and the next set of test data is used. If not, the relevant parameters of the AI ​​model need to be adjusted, and the test data set is used again for testing. This process continues until a predetermined proportion of the test data passes the test. The final result is an intelligent functional model whose input is multi-source features and their corresponding feature function scores, and whose output is the feature function score.

[0008] Optionally, the aggregation optimization process includes: The association similarity is calculated by examining the multi-source features stored at the same level in any two storage units; the specific formula for calculating the association similarity is as follows: in, This represents the association similarity between the i-th and j-th storage units at level l. This represents the multi-source feature vector of the i-th storage unit in the l-th layer. This represents the multi-source feature vector of the j-th storage unit in the l-th layer; and They represent and The corresponding multi-source features, where k represents the number of types of multi-source features, i≠j≤N; l≤L, N represents the total number of storage units, and L represents the total number of levels; This represents the dot product operation between vectors. and These represent the product of the Euclidean lengths.

[0009] Obtain all association similarities at the same level, and compare each association similarity with the association similarity threshold in turn; if the association similarity is less than the association similarity threshold, the association similarity is determined to be an abnormal aggregation, an abnormal label is generated and transmitted to the control terminal; The corrected association similarity is obtained by performing anomaly correction on the association similarity of abnormal tags through the control terminal; the association similarity of those not judged as abnormal aggregations and the corrected association similarity are integrated to obtain the first association similarity set; The aggregate credibility of each storage unit at the same level is calculated based on the first association similarity set. The specific formula for calculating the aggregation credibility is as follows: in, This represents the aggregation confidence level of the i-th storage unit at layer l.

[0010] Optionally, the step of filtering out abnormal features of multi-source features based on the aggregated credibility of all levels of each storage unit to obtain the first multi-source feature includes: sequentially comparing the aggregated credibility of each storage unit at the same level with the credibility threshold. When the aggregate confidence level is less than the confidence level threshold, the multi-source features of the current comparison storage unit will be filtered. When the aggregation confidence level is greater than or equal to the confidence level threshold, the multi-source features of the current comparison storage unit are retained. Obtain all retained multi-source features, and sequentially perform convolution on all retained multi-source features to obtain multiple first convolutional features; Multiple first convolutional features are subjected to point-like convolution to obtain multiple second convolutional features with the same number of feature channels; The first multi-source feature is obtained by fusing the multiple second convolutional features; The calculation process for the credibility threshold includes: in, Let represent the confidence threshold at level l, where i and j represent any one of the storage units, i ≠ j ≤ N; l ≤ L, where L represents the total number of levels and N represents the total number of storage units. This represents the aggregation confidence level of the i-th storage unit at layer l.

[0011] Optionally, the aggregation weight is obtained by weighted aggregation calculation based on the aggregation credibility, and the aggregation feature is obtained by weighted averaging of the aggregation weight and the first multi-source feature, including: The calculation process for the aggregation weight is as follows: in, This represents the aggregation confidence level of the i-th storage unit after filtering at layer l, and N represents the total number of storage units; The formula for calculating the aggregation feature is: in, Let A represent the clustering feature, and let A represent the first multi-source feature.

[0012] In a second aspect of this invention, a multi-source data aggregation system for implementing hierarchical querying is proposed, the system comprising: Data acquisition module: Acquires multi-source data from various storage units of the target enterprise, and performs feature generation based on the multi-source data to obtain multi-source features; Intelligent function module: The feature function score is obtained by substituting the multi-source features into the intelligent function model; Aggregation optimization module: Divides multi-source features into multiple levels based on their data importance through feature function scoring; aggregates and optimizes multi-source features of the same level in each storage unit to obtain the aggregation credibility of each storage unit; Anomaly filtering module: Based on the aggregation credibility of each storage unit, the multi-source features are filtered for anomalies to obtain the first multi-source feature; The aggregation feature module calculates the aggregation weight based on the aggregation credibility, and calculates the aggregation feature by averaging the aggregation weight and the first multi-source feature. Aggregation result generation module: Generates aggregation results based on aggregation features; obtains the aggregation results of each storage unit at all levels.

[0013] Optionally, the intelligent function module is further used for: Obtain multi-source features and feature function scores from the database; integrate the multi-source features and their corresponding feature function scores into several training and testing data sets; Several training data are imported into the artificial intelligence model for training, and the trained artificial intelligence model is tested with test data; finally, an intelligent functional model is obtained with multi-source features and their corresponding feature function scores as inputs and feature function scores as outputs.

[0014] Optionally, the aggregation optimization module includes: an association similarity module, a judgment module, a first association similarity set module, and an aggregation credibility module. The association similarity module is used to calculate the association similarity of multi-source features stored at the same level in any two storage units; the specific calculation formula for the association similarity is as follows:

[0015] in, This represents the association similarity between the i-th and j-th storage units at level l. This represents the multi-source feature vector of the i-th storage unit in the l-th layer. This represents the multi-source feature vector of the j-th storage unit in the l-th layer; and They represent and The corresponding multi-source features, where k represents the number of types of multi-source features, i≠j≤N; l≤L, N represents the total number of storage units, and L represents the total number of levels; This represents the dot product operation between vectors. and These represent the product of the Euclidean lengths.

[0016] The judgment module is used to obtain all association similarities at the same level, and compare all association similarities with the association similarity threshold in turn; if the association similarity is less than the association similarity threshold, the association similarity is judged as an abnormal aggregation, an abnormal label is generated and transmitted to the control terminal. The first association similarity set module is used to perform abnormal repair on the association similarity of abnormal tags through the control terminal to obtain repaired association similarity; and to integrate the association similarity that is not judged as abnormal aggregation and the repaired association similarity to obtain the first association similarity set; The aggregation credibility module is used to calculate the aggregation credibility of each storage unit at the same level based on the first association similarity set. The specific formula for calculating the aggregation credibility is as follows: in, This represents the aggregation confidence level of the i-th storage unit at layer l.

[0017] Optionally, the anomaly feature filtering module includes: a comparison module, a first convolution module, a second convolution module, and a first multi-source feature. The comparison module is used to compare the aggregate confidence and confidence threshold of each storage unit at the same level in turn; when the aggregate confidence is less than the confidence threshold, the multi-source features of the currently compared storage unit are filtered out; when the aggregate confidence is greater than or equal to the confidence threshold, the multi-source features of the currently compared storage unit are retained. The first convolution module is used to obtain all retained multi-source features, and to sequentially convolve all retained multi-source features to obtain multiple first convolution features; The second convolution module is used to perform point convolution on multiple first convolution features to obtain multiple second convolution features with the same number of feature channels; The first multi-source feature module is used to fuse the multiple second convolutional features to obtain the first multi-source feature; The calculation process for the credibility threshold includes: in, Let represent the confidence threshold at level l, where i and j represent any one of the storage units, i ≠ j ≤ N; l ≤ L, where L represents the total number of levels and N represents the total number of storage units. This represents the aggregation confidence level of the i-th storage unit at layer l.

[0018] Optionally, the aggregation feature module is further configured to: The calculation process for the aggregation weight is as follows: in, This represents the aggregation confidence level of the i-th storage unit after filtering at layer l, and N represents the total number of storage units; The formula for calculating the aggregation feature is: in, Let A represent the clustering feature, and let A represent the first multi-source feature.

[0019] The beneficial effects of this invention are: This invention proposes a method and system for hierarchical querying of multi-source data aggregation. By classifying multi-source data in each storage unit of the target enterprise according to the hierarchy of data importance, and then performing aggregation optimization on multi-source data at the same level, it overcomes the inherent defects of multi-source data stratification within storage units and achieves a dual improvement in the efficiency and accuracy of multi-source data aggregation. Attached Figure Description

[0020] The invention will now be further described with reference to the accompanying drawings.

[0021] Figure 1 A flowchart illustrating a multi-source data aggregation method for implementing hierarchical querying, provided by an embodiment of the present invention; Figure 2 A flowchart for aggregation optimization provided in an embodiment of the present invention; Figure 3 A data flow diagram for anomaly feature filtering provided in an embodiment of the present invention; Figure 4 This is a framework diagram of a multi-source data aggregation system for implementing hierarchical querying, provided by an embodiment of the present invention. Detailed Implementation

[0022] 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. The term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and B can represent: A alone, A and B simultaneously, and B alone. Furthermore, descriptions involving "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined with "first" or "second" can explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0023] 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.

[0024] This invention provides a method for hierarchical querying of multi-source data aggregation. See also... Figure 1 , Figure 1 This is a flowchart illustrating a method for hierarchical querying and multi-source data aggregation according to an embodiment of the present invention. The method includes the following steps: Acquire multi-source data from various storage units of the target enterprise, and perform feature extraction based on the multi-source data to obtain multi-source features; The feature function score is obtained by substituting multi-source features into the intelligent function model; The data importance hierarchy of multi-source features is divided into multiple levels by feature function scoring; the aggregation and optimization of multi-source features of any level in each storage unit is performed to obtain the aggregation credibility of each storage unit. The first multi-source feature is obtained by filtering out abnormal features based on the aggregation credibility of each storage unit. The aggregation weight is obtained by weighted aggregation calculation based on the aggregation credibility, and the aggregation feature is obtained by weighted averaging of the aggregation weight and the first multi-source feature. The aggregation result is generated based on the aggregation characteristics; Obtain the aggregated results of each storage unit across all levels; Based on the multi-source data aggregation method for hierarchical query provided by the embodiments of the present invention, the multi-source data of each storage unit of the target enterprise is divided into multiple levels according to the importance level of the data, and the aggregation optimization is performed on the multi-source data of any same level. This solves the defect of hierarchical multi-source data within the storage unit and improves the aggregation efficiency and accuracy of multi-source data.

[0025] In one implementation, multi-source features are obtained by characterizing multi-source data. Specifically, the multi-source data is cleaned and then standardized. After that, features are extracted from structured and unstructured data, and finally, features are fused to generate multi-source features. Specifically, the multi-source data includes sales data, cost data, and profit data, etc. In one implementation, the training process of the intelligent functional model includes: Multi-source features and feature function scores are obtained from the database. The multi-source features include sales features, cost features, and profit features. The feature function scores are obtained by experts through integrated functional evaluation based on the multi-source features. The intelligent functional model is a functional layer consisting of standard convolutional layers, convolutional and fully connected transition layers, and fully connected layers. The multi-source features and their corresponding feature function scores are integrated into several training data and test data.

[0026] Several training datasets are imported into the AI ​​model for training, and then the trained AI model is tested using test data. Specifically, the multi-source features from the test data are input into the trained AI model, and the output is a feature function score. The absolute value of the difference between the feature function score and the feature function score recorded in the test data is checked to see if it is within an acceptable range. If yes, the test data set passes the test, and the next set of test data is used for testing. If not, the relevant parameters of the AI ​​model need to be adjusted, and the test data set is used again for testing. This continues until a set proportion of the test data passes the test. Finally, an intelligent functional model is obtained, with multi-source features and their corresponding feature function scores as inputs and feature function scores as outputs. The AI ​​model is a CNN+MLP model.

[0027] In one implementation, the specific structure of the artificial intelligence model is as follows: Standard convolutional layer: a 1D convolutional layer (Conv1D) is used, with the input dimension adapted to a 64-dimensional multi-source feature vector, 32 kernels, kernel size 3, stride 1, ReLU activation function, and same padding method, used to extract local correlation patterns of features; Convolutional and fully connected transition layer: a global average pooling layer (GAP) reduces the dimensionality of the convolutional output to 32 dimensions, followed by a batch normalization layer (BatchNorm) to eliminate dimensionality differences and prevent overfitting; Fully connected layer: the first layer has 128 neurons (ReLU activation), the second layer has 64 neurons (ReLU activation), and the output layer has 1 neuron (linear activation), outputting a feature function score. The optimizer used was Adam, with a learning rate of 0.001, β1=0.9, β2=0.998, and weight decay of 1e-5 to balance convergence speed and stability. Training parameters included a batch size of 32, 50 training epochs, and mean squared error (MSE) as the loss function. An early stopping strategy (patience=5) was set to avoid overfitting. The test threshold was set to ±0.05 (out of 1) for the absolute value of the difference in feature function scores, and training was considered complete when more than 85% of the test data passed.

[0028] In one implementation, see [link to implementation details]. Figure 2 , Figure 2 A flowchart for aggregation optimization is provided as an embodiment of the present invention. The aggregation optimization process includes: Step 1: Calculate the association similarity for multi-source features stored at the same level in any two storage units; the specific formula for calculating the association similarity is: in, This represents the association similarity between the i-th and j-th storage units at level l. This represents the multi-source feature vector of the i-th storage unit in the l-th layer. This represents the multi-source feature vector of the j-th storage unit in the l-th layer; and They represent and The corresponding multi-source features, where k represents the number of types of multi-source features, i≠j≤N; l≤L, N represents the total number of storage units, and L represents the total number of levels; This represents the dot product operation between vectors. and These represent the product of the Euclidean lengths.

[0029] Step 2: Obtain all association similarities at the same level, and compare each association similarity with the association similarity threshold in turn; if the association similarity is less than the association similarity threshold, the association similarity is determined to be an abnormal aggregation, an abnormal label is generated and transmitted to the control terminal; Step 3: The association similarity of abnormal tags is repaired by controlling the terminal to obtain the repaired association similarity; the association similarity that is not judged as abnormal aggregation and the repaired association similarity are integrated to obtain the first association similarity set; Step 4: Calculate the aggregate credibility of each storage unit at the same level based on the first association similarity set; The specific formula for calculating aggregate credibility is as follows: in, This represents the aggregation confidence level of the i-th storage unit at layer l.

[0030] In one implementation, the layers can specifically be functional layers such as standard convolutional layers, convolutional-fully connected transition layers, and fully connected layers. By calculating the cosine correlation similarity of multi-source features at the hierarchical level of storage units, combined with anomaly detection and repair and aggregation credibility assessment, the accuracy and reliability of storage system aggregation optimization can be significantly improved. For example, in the standard convolutional layer of the intelligent construction standard system identification task, calculating the correlation similarity of feature storage units can quickly locate abnormally aggregated storage unit pairs. After repair, these pairs are integrated to generate a highly credible correlation similarity set. This effectively reduces redundant storage overhead while enhancing the system's ability to perceive multi-source feature correlations, providing a quantitative basis for intelligent decision-making in the storage system, and ensuring the stability and efficiency of data aggregation.

[0031] In one implementation, see [link to implementation details]. Figure 3 , Figure 3 This invention provides a data flow graph for anomaly feature filtering. The first multi-source feature is obtained by filtering multi-source features based on the aggregated confidence level of all levels of each storage unit, including: The aggregation confidence level and confidence threshold of each storage unit at the same level are compared sequentially. When the aggregate confidence level is less than the confidence level threshold, the multi-source features of the current comparison storage unit will be filtered. When the aggregation confidence level is greater than or equal to the confidence level threshold, the multi-source features of the current comparison storage unit are retained. Obtain all retained multi-source features, and sequentially convolve all retained multi-source features to obtain multiple first convolutional features; Multiple first convolutional features are subjected to point-like convolution to obtain multiple second convolutional features with the same number of feature channels; The first multi-source feature is obtained by fusing multiple second convolutional features. The calculation process for the credibility threshold includes: in, Let represent the confidence threshold at level l, where i and j represent any one of the storage units, i ≠ j ≤ N; l ≤ L, where L represents the total number of levels and N represents the total number of storage units. This represents the aggregation confidence level of the i-th storage unit at layer l.

[0032] In one implementation, the credibility threshold is calculated by averaging the maximum aggregation credibility within a level, which effectively avoids interference from extreme values ​​and improves the accuracy and robustness of abnormal feature filtering. By comparing the aggregation credibility with the threshold, abnormal and noisy features with low credibility are accurately removed, improving the quality of input features from the source. Subsequently, convolution and point convolution are performed on the retained features to unify the number of channels and feature fusion, which not only strengthens the local representation ability and nonlinear expression ability of the features, but also optimizes the consistency of feature dimensions. The first multi-source feature generated in the end integrates multi-dimensional effective information.

[0033] In one implementation, the aggregation weight is obtained by weighted aggregation calculation based on the aggregation credibility, and the aggregated features are obtained by weighted averaging of the aggregation weight and the first multi-source features, including: The calculation process for the aggregate weight is as follows: in, This represents the aggregation confidence level of the i-th storage unit after filtering at layer l, and N represents the total number of storage units; The formula for calculating aggregation features is: in, Let A represent the clustering feature, and let A represent the first multi-source feature.

[0034] In one implementation, aggregation weights are calculated using normalized calculations based on the aggregated credibility after filtering. This ensures that weight allocation is positively correlated with the credibility of the storage unit features, with higher credibility features receiving higher weights and lower credibility features receiving lower weights. This mechanism precisely strengthens the influence of effective features while mitigating the interference of noise and low-quality features, making weight allocation more targeted and reasonable. The subsequent process of combining the weights with the first multi-source features to obtain the aggregated features preserves the diversity of multi-source features and ensures that high-credibility features dominate the final aggregation result through the weighting mechanism, reducing the negative impact of low-quality features on the overall representation. The entire process, from weight allocation to feature aggregation, revolves around credibility optimization, effectively improving the quality and representational ability of the aggregated features, thereby enhancing the accuracy and robustness of the model, increasing aggregation efficiency, and improving the efficiency and accuracy of subsequent user queries.

[0035] Based on the same inventive concept, embodiments of the present invention also provide a multi-source data aggregation system for implementing hierarchical querying. See also Figure 4 , Figure 4 A framework diagram of a multi-source data aggregation system for implementing hierarchical querying, provided in an embodiment of the present invention, includes: Data acquisition module: Acquires multi-source data from various storage units of the target enterprise, and performs feature generation based on the multi-source data to obtain multi-source features; Intelligent Function Module: The feature function score is obtained by substituting multi-source features into the intelligent function model; Aggregation optimization module: Divides multi-source features into multiple levels based on their data importance through feature function scoring; aggregates and optimizes multi-source features of the same level in each storage unit to obtain the aggregation credibility of each storage unit; Anomaly filtering module: Based on the aggregation credibility of each storage unit, the multi-source features are filtered for anomalies to obtain the first multi-source feature; The aggregation feature module calculates the aggregation weight based on the aggregation credibility, and then calculates the aggregation feature by averaging the aggregation weight and the first multi-source feature. Aggregation result generation module: Generates aggregation results based on aggregation features; obtains the aggregation results of each storage unit at all levels.

[0036] Based on the embodiments of the present invention, a multi-source data aggregation system for hierarchical query is provided. By dividing the multi-source data of each storage unit of the target enterprise into multiple levels according to the importance level of the data, and performing aggregation optimization on multi-source data of any level, the system solves the defect of hierarchical multi-source data within the storage unit and improves the aggregation efficiency and accuracy of multi-source data.

[0037] The foregoing has described one embodiment of the present invention in detail, but this content is merely a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the scope of the claims of this invention.

Claims

1. A method for multi-source data aggregation to achieve hierarchical querying, characterized in that, The method includes: Acquire multi-source data from each storage unit of the target enterprise, and perform feature extraction based on the multi-source data to obtain multi-source features; The feature function score is obtained by substituting the multi-source features into the intelligent function model; The data importance hierarchy of multi-source features is divided into multiple levels by feature function scoring; the aggregation and optimization of multi-source features of any level in each storage unit is performed to obtain the aggregation credibility of each storage unit. The first multi-source feature is obtained by filtering out abnormal features based on the aggregation credibility of each storage unit. The aggregation weight is obtained by weighted aggregation calculation based on the aggregation credibility, and the aggregation feature is obtained by weighted averaging of the aggregation weight and the first multi-source feature. The aggregation result is generated based on the aggregation characteristics; Obtain the aggregated results of each storage unit across all levels.

2. The method for hierarchical querying of multi-source data aggregation according to claim 1, characterized in that, The training process of the intelligent functional model includes: Obtain multi-source features and feature function scores from the database; integrate the multi-source features and their corresponding feature function scores into several training and testing data sets; Several training data are imported into the artificial intelligence model for training, and the trained artificial intelligence model is tested with test data; finally, an intelligent functional model is obtained with multi-source features and their corresponding feature function scores as inputs and feature function scores as outputs.

3. The method for multi-source data aggregation to achieve hierarchical querying according to claim 1, characterized in that, The aggregation optimization process includes: The correlation similarity is calculated for the multi-source features of the same storage level of any two storage units; Obtain all association similarities at the same level, and compare each association similarity with the association similarity threshold in turn; if the association similarity is less than the association similarity threshold, the association similarity is determined to be an abnormal aggregation, an abnormal label is generated and transmitted to the control terminal; The corrected association similarity is obtained by performing anomaly correction on the association similarity of abnormal tags through the control terminal; the association similarity of those not judged as abnormal aggregations and the corrected association similarity are integrated to obtain the first association similarity set; The aggregate credibility of each storage unit at the same level is calculated based on the first association similarity set.

4. The method for multi-source data aggregation to achieve hierarchical querying according to claim 1, characterized in that, The step of filtering out abnormal features from the multi-source features based on the aggregated credibility of all levels of each storage unit to obtain the first multi-source feature includes: The aggregation confidence level and confidence threshold of each storage unit at the same level are compared sequentially. When the aggregate confidence level is less than the confidence level threshold, the multi-source features of the current comparison storage unit will be filtered. When the aggregation confidence level is greater than or equal to the confidence level threshold, the multi-source features of the current comparison storage unit are retained. Obtain all retained multi-source features, and sequentially perform convolution on all retained multi-source features to obtain multiple first convolutional features; Multiple first convolutional features are subjected to point-like convolution to obtain multiple second convolutional features with the same number of feature channels; The first multi-source feature is obtained by fusing the multiple second convolutional features.

5. The method for multi-source data aggregation to achieve hierarchical querying according to claim 1, characterized in that, The aggregation weight is obtained by weighted aggregation calculation based on the aggregation credibility, and the aggregation features are obtained by weighted averaging of the aggregation weight and the first multi-source features, including: The calculation process for the aggregation weight is as follows: in, This represents the aggregation confidence level of the i-th storage unit after filtering at layer l, and N represents the total number of storage units; The formula for calculating the aggregation feature is: in, Let A represent the clustering feature, and let A represent the first multi-source feature.

6. A multi-source data aggregation system for implementing hierarchical querying, characterized in that, The system includes: Data acquisition module: Acquires multi-source data from various storage units of the target enterprise, and performs feature generation based on the multi-source data to obtain multi-source features; Intelligent function module: The feature function score is obtained by substituting the multi-source features into the intelligent function model; Aggregation optimization module: Divides multi-source features into multiple levels based on their data importance through feature function scoring; aggregates and optimizes multi-source features of the same level in each storage unit to obtain the aggregation credibility of each storage unit; Anomaly filtering module: Based on the aggregation credibility of each storage unit, the multi-source features are filtered for anomalies to obtain the first multi-source feature; The aggregation feature module calculates the aggregation weight based on the aggregation credibility, and calculates the aggregation feature by averaging the aggregation weight and the first multi-source feature. Aggregation result generation module: Generates aggregation results based on aggregation features; obtains the aggregation results of each storage unit at all levels.

7. A multi-source data aggregation system for implementing hierarchical querying according to claim 6, characterized in that, The intelligent function module is also used for: Obtain multi-source features and feature function scores from the database; integrate the multi-source features and their corresponding feature function scores into several training and testing data sets; Several training data are imported into the artificial intelligence model for training, and the trained artificial intelligence model is tested with test data; finally, an intelligent functional model is obtained with multi-source features and their corresponding feature function scores as inputs and feature function scores as outputs.

8. A multi-source data aggregation system for implementing hierarchical querying according to claim 6, characterized in that, The aggregation optimization module includes: an association similarity module, a judgment module, a first association similarity set module, and an aggregation credibility module. The association similarity module is used to calculate the association similarity of multi-source features stored at the same level in any two storage units. The judgment module is used to obtain all association similarities at the same level, and compare all association similarities with the association similarity threshold in turn; if the association similarity is less than the association similarity threshold, the association similarity is judged as an abnormal aggregation, an abnormal label is generated and transmitted to the control terminal; The first association similarity set module is used to perform abnormal repair on the association similarity of abnormal tags through the control terminal to obtain repaired association similarity; and to integrate the association similarity that is not judged as abnormal aggregation and the repaired association similarity to obtain the first association similarity set; The aggregation credibility module is used to calculate the aggregation credibility of each storage unit at the same level based on the first association similarity set.

9. A multi-source data aggregation system for implementing hierarchical querying according to claim 6, characterized in that, The abnormal feature filtering module includes: a comparison module, a first convolution module, a second convolution module, and a first multi-source feature. The comparison module is used to compare the aggregate confidence and confidence threshold of each storage unit at the same level in turn; when the aggregate confidence is less than the confidence threshold, the multi-source features of the storage unit being compared are filtered out; when the aggregate confidence is greater than or equal to the confidence threshold, the multi-source features of the storage unit being compared are retained. The first convolution module is used to obtain all retained multi-source features, and to sequentially convolve all retained multi-source features to obtain multiple first convolution features; The second convolution module is used to perform point convolution on multiple first convolution features to obtain multiple second convolution features with the same number of feature channels; The first multi-source feature module is used to perform feature fusion on the multiple second convolutional features to obtain the first multi-source feature.

10. A multi-source data aggregation system for implementing hierarchical querying according to claim 6, characterized in that, The aggregation feature module is also used for: The calculation process for the aggregation weight is as follows: in, This represents the aggregation confidence level of the i-th storage unit after filtering at layer l, and N represents the total number of storage units; The formula for calculating the aggregation feature is: in, Let A represent the clustering feature, and let A represent the first multi-source feature.

Images