A dynamically configurable BI multidimensional analysis report generation system

Abstract

The application relates to a dynamically configurable BI multidimensional analysis report generation system and belongs to the technical field of intelligent business data analysis. In the application, a data access and logic modeling unit acquires a multi-source original business data set, performs data type standardization processing, and constructs a multidimensional logic model; a semantic mapping and index configuration unit constructs a virtual semantic layer to translate fields into business semantic entities, generates a dimension and measurement set, performs calculation measurement rule binding on a measurement item, and outputs a semantic configuration scheme; a multidimensional filtering and instant query unit acquires a multidimensional filtering condition group and a graphic display type, performs filtering matching and multidimensional association combination on the dimension and measurement set, performs multidimensional instant query to generate a multidimensional result data set; and a report assembly and visualized publishing unit matches a visualized chart template, performs data binding and layout filling on the multidimensional result data set and the template, and generates a multidimensional analysis report. The application realizes dynamic configurable closed-loop generation from multi-source data to a multidimensional analysis report.

Description

Technical Field

[0001] This invention belongs to the field of intelligent business data analysis technology, specifically relating to a dynamically configurable BI multidimensional analysis report generation system. Background Technology

[0002] As enterprises accelerate their digital transformation, data is scattered across multiple heterogeneous data sources such as ERP, CRM, and financial systems. How to quickly integrate this data and generate multidimensional analytical reports has become a key challenge for enterprises in making data-driven decisions.

[0003] Existing BI report generation systems typically employ ETL processes to extract, clean, and transform data from multiple sources before loading it into a data warehouse. This approach relies on predefined data models and fixed ETL workflows. When business requirements change, technical personnel need to redesign the data extraction process, adjust the warehouse table structure, and rebuild the data model, resulting in long report generation cycles and delayed responses. Furthermore, for cross-business report integration needs, existing solutions generally lack conflict resolution mechanisms when handling mixed storage of multi-source data. Inconsistency in field naming, data types, and encoding methods across different data sources is difficult to guarantee, leading to data redundancy and accuracy issues in subsequent report generation.

[0004] At the semantic mapping level, while traditional BI tools offer semantic modeling capabilities, a natural gap exists between the technical field names stored in the physical database and the terminology used daily by business personnel. In the database, fields might be "sale_amt" or "revenue_amount," while business personnel are accustomed to using business concepts such as "sales revenue" or "income." Existing solutions mostly rely on technical personnel manually writing mapping configurations to map physical fields one by one to business aliases, failing to achieve automated translation from physical fields to business semantic entities. When the database table structure changes or new data sources are added, the workload of maintaining the mapping configuration increases linearly.

[0005] At the report filtering and query level, users typically need to manually associate dimensions and measures by dragging and dropping or clicking, which is inefficient. Some solutions utilize indicator-dimensional matrices to achieve dynamic report generation, using matrix operations to filter and merge dimensions and indicators, thus improving report generation efficiency. However, the dimension-indicator matrix on which these solutions rely is statically preset, and the binding relationship between dimensions and measures is determined during the configuration phase. When the user's filtering conditions include new dimension combinations or customized measure calculation requirements, it is still necessary to return to the configuration page to redefine them, lacking the ability to adaptively match semantic solutions based on the user's dynamic filtering conditions.

[0006] In the report assembly and publishing phase, visual charts typically use preset templates for data rendering. When configuring reports, users need to manually map the fields in the result dataset to each data binding point in the chart component. When the field structure of the result dataset changes due to changes in filtering conditions, the preset data binding points in the template struggle to adapt to the new field structure, leading to misaligned data bindings or rendering errors. Users are forced to repeatedly adjust the template configuration. Furthermore, report publishing is usually geared towards a single output platform, requiring manual format adaptation for different devices, making the publishing process cumbersome and difficult to track in a closed-loop manner.

[0007] In summary, existing BI report generation systems have not yet achieved a closed-loop end-to-end process, from multi-source data access, automatic semantic mapping, dynamic linkage of filtering conditions and semantic schemes to adaptive report assembly and publishing, making it difficult to meet enterprises' needs for dynamic configurability and real-time response in report generation. Summary of the Invention

[0008] To address the aforementioned problems in the existing technology, this invention provides a dynamically configurable BI multidimensional analysis report generation system. The objective of this invention can be achieved through the following technical solutions: The data access and logical modeling unit acquires the original business dataset from multiple data source modules, extracts the field attributes and inter-table relationships of each data table in the original business dataset, performs data type standardization processing on the field attributes, and constructs a multi-dimensional logical model based on the inter-table relationships; The semantic mapping and indicator configuration unit extracts the physical name and data type of each field in the multidimensional logical model, constructs a virtual semantic layer, translates each field into the corresponding business semantic entity through the virtual semantic layer, generates a dimension set and a measurement set, calculates measurement rules and binds each measurement item in the measurement set, and outputs a multidimensional analysis semantic configuration scheme. The multidimensional filtering and real-time query unit obtains the multidimensional filtering condition group and graphic display type input by the report user, performs filtering matching and multidimensional association combination on the dimension set and metric set in the multidimensional analysis semantic configuration scheme based on the multidimensional filtering condition group, performs multidimensional real-time query based on the association combination result, and generates a multidimensional result dataset. The report assembly and visualization publishing unit acquires the multidimensional result dataset, matches the corresponding visualization chart template based on the graphic display type, performs data binding and layout filling on the multidimensional result dataset and the visualization chart template, and generates a multidimensional analysis report.

[0009] Specifically, the data type standardization process includes: Establish a cross-data source data type mapping benchmark, match the original data type identifiers of each field in different data sources with the data type mapping benchmark, and generate a standardized data type mapping table; Perform format compliance validation on the original data values ​​of each field, and correct any abnormal data values ​​that fail the validation according to the preset abnormal data processing rules; The corrected data values ​​are formatted according to the corresponding standardized data type identifier to generate standardized field data.

[0010] Specifically, the construction process of the multidimensional logical model includes: Extract the primary key and foreign key identifiers of each data table. Based on the matching relationship between the primary key and foreign key identifiers, generate a central fact table. Then, identify the data tables referenced by the central fact table through foreign key associations as the corresponding fact dimension tables. The central fact table is used as the central node of the star-shaped structure of the multidimensional logical model, and each fact dimension table is linked to the peripheral nodes of the central fact table through foreign key associations. Perform fact-dimensional reference validation on the completed multidimensional logical model. After the validation passes, output the final multidimensional logical model.

[0011] Specifically, the construction process of the virtual semantic layer includes: Extract the physical name and data type of each field in the multidimensional logical model, match the physical name with a preset business term dictionary, and generate business semantic entities corresponding to each field; Write the translation mapping relationship between physical fields and business semantic entities into the mapping structure of the virtual semantic layer to establish a bidirectional translation link between the physical layer and the semantic layer. When multiple candidate semantic entity identifiers exist for the same physical field, semantic disambiguation is performed. A unique translation mapping relationship is determined based on the business context relevance of the semantic entity identifier, and a virtual semantic layer is constructed.

[0012] Specifically, the multidimensional analysis semantic configuration scheme includes: Through the bidirectional translation link of the virtual semantic layer, the physical fields in the multidimensional logical model are translated into corresponding business semantic entities; Entities with discrete attributes in the business semantic entity are classified as dimensions, generating a dimension set; entities with continuous numerical values ​​in the business semantic entity are classified as measures, generating a measure set. A preset dynamic binding mechanism for measurement rules is invoked for each measurement item in the measurement set. The measurement rule binding is performed on the measurement item through the dynamic binding mechanism, and a multidimensional analysis semantic configuration scheme is output.

[0013] Specifically, the dynamic binding mechanism for measurement rules includes: Perform a multi-dimensional cross-combination of the dimensions in the dimension set and the measures in the measure set to generate a candidate set of measure items; The metrics that meet the preset dimension level derivation conditions in the metric candidate set are marked as derived metrics, and calculation metric rules are generated based on the dimension combination and aggregation method corresponding to the derived metrics. The generated computational metric rules are bound to the corresponding derived metric items to generate a complete set of metric rule bindings.

[0014] Specifically, the combination of filtering and matching with multidimensional association includes: The filter condition elements corresponding to each filter condition in the multidimensional filter condition group are parsed, and the parsed filter condition elements are dynamically mapped and matched with the dimension set and metric set in the multidimensional analysis semantic configuration scheme to generate a filter matching mapping table. Extract the logical operators between each filtering condition in the filtering matching mapping table, and combine the matched dimension fields and metric fields into a multi-dimensional association structure based on the logical operators to generate a dimension-metric association graph.

[0015] Specifically, the dynamic mapping matching includes: Extract the target dimension field from the filtering condition elements, perform semantic similarity matching between the target dimension field and the business semantic entity in the virtual semantic layer, and determine the corresponding dimension field. Extract the condition values ​​and comparison operators from the filtering condition elements, verify the legality of the condition values ​​and comparison operators based on the semantic range of the dimension field, and establish a filtering binding relationship between the verified condition values ​​and the corresponding business semantic entities in the virtual semantic layer. For condition values ​​that fail validation, semantic re-parsing is triggered. Based on the re-parsing result, the binding relationship is updated, and a complete filtering and matching mapping table is generated.

[0016] Specifically, the generation process of the dimension-metric association graph includes: Based on the logical operators between the filtering conditions in the filtering matching mapping table, the matched dimension fields and metric fields are associated according to the connection rules corresponding to the logical operators to generate an initial association structure. Perform association strength calculation on the connection edges between dimension nodes and metric nodes in the initial association structure to form a set of dimension-metric association edges; By performing connectivity checks on the dimension-metric association edge set, nodes with connected paths are combined as connected subgraphs to generate the dimension-metric association graph.

[0017] Specifically, the multidimensional real-time query includes: The dimension-metric association graph is deconstructed into a set of dimension nodes and a set of metric nodes. The computational metric rule references of each metric node in the metric node set are extracted, and a parsing binding between the metric node and the computational metric rule is established. Based on the hierarchical relationship of the dimension nodes in the dimension node set and the inter-table relationship in the multidimensional logical model, semantic reconstruction is performed on the parsed and bound metric nodes to generate a multidimensional query intermediate structure. The metric reference expansion operation is performed on the metric nodes that reference the metric calculation rules in the multidimensional query intermediate structure. The expanded metric references are then assembled with the corresponding aggregation operations into a multidimensional aggregation operation queue to generate a multidimensional result dataset.

[0018] Specifically, the data binding and layout filling include: Perform hierarchical parsing on the preset data binding points of each chart component in the visualization chart template to determine the dimension level and metric aggregation granularity corresponding to each data binding point; The data in the multidimensional result dataset are grouped and adapted according to the dimension hierarchy and metric aggregation granularity. The adapted data groups are then mapped to the corresponding data binding points for hierarchical data binding. After data binding, each chart component is adaptively filled according to the layout structure of the visualization chart template. After filling, the inter-component linkage relationship is established for each chart component, and a multi-dimensional analysis report is generated.

[0019] Specifically, the data grouping adaptation includes: Extract the required dimensional hierarchy and metric aggregation granularity for each data binding point, and match the aggregation granularity of the multidimensional result dataset with the required granularity of the data binding points; For data whose aggregation granularity does not match the granularity required by the data binding point, perform granularity re-aggregation operation to aggregate the data to the granularity level required by the data binding point; Data is grouped layer by layer from coarse-grained to fine-grained according to the dimensional hierarchy, and a hierarchical mapping and binding relationship is established between each group of data and the corresponding data binding point at the granularity level.

[0020] The beneficial effects of this invention are as follows: (1) By constructing a collaborative architecture of data access and logical modeling unit and semantic mapping and indicator configuration unit, the technical problem of delayed report generation response caused by the reliance on preset ETL processes and fixed data models in the existing multi-source data integration is solved. The data access and logical modeling unit establishes a cross-data source data type mapping benchmark, performs standardized processing on the original data type identifiers of each field in different data sources, and automatically identifies fact tables and dimension tables through the matching relationship of primary keys and foreign keys. Based on the star structure, it constructs a multi-dimensional logical model, realizing the automated integration of multi-source heterogeneous data. The semantic mapping and indicator configuration unit constructs a virtual semantic layer, translates physical fields into corresponding business semantic entities, and performs semantic disambiguation processing on the case where there are multiple candidate semantic entity identifiers for the same physical field, realizing the automated mapping between the physical layer and the semantic layer. At the same time, by triggering the dynamic binding mechanism of measurement rules, the dimension set and the measurement set are combined in a multi-dimensional way, automatically generating derived measurement items and executing the calculation measurement rule binding, so that the generation of measurement items no longer depends on manual definition by technical personnel, improving the automation level and dynamic configurability of report generation.

[0021] (2) By constructing a closed-loop architecture of multidimensional filtering and real-time query unit and report assembly and visualization publishing unit, the technical problems of insufficient linkage between filtering conditions and semantic scheme and difficulty in adaptively matching field structure changes of data binding points in visualization reports in the existing solution are solved. The multidimensional filtering and real-time query unit establishes the filtering binding relationship between filtering conditions and semantic layer business semantic entities by performing dynamic mapping matching between the parsed filtering condition elements and the dimension set and metric set in the multidimensional analysis semantic configuration scheme, and triggers semantic re-parsing to update the binding relationship for the condition values ​​that fail the verification, thus realizing the dynamic linkage between filtering conditions and semantic scheme; in the query stage, the dimension-metric association graph is deconstructed into a node set, and a multidimensional query intermediate structure is generated through semantic reconstruction. Then, the metric reference expansion operation is performed on the metric nodes that reference the metric calculation rules to generate a multidimensional result dataset. The report assembly and visualization publishing unit performs hierarchical parsing on the data binding points of chart components, performs granular re-aggregation operations on data with mismatched aggregation granularity, and establishes hierarchical mapping binding relationships layer by layer according to the dimension level. This enables the report to automatically adjust the data filling granularity and layout structure when the filtering conditions change, realizing the full-link dynamic configurable generation of multidimensional analysis reports. Attached Figure Description

[0022] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0023] Figure 1 This is a flowchart illustrating a dynamically configurable BI multidimensional analysis report generation system according to the present invention. Figure 2This is a timing diagram of a dynamically configurable BI multidimensional analysis report generation system according to the present invention. Detailed Implementation

[0024] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0025] Please see Figures 1-2 A dynamically configurable BI multidimensional analysis report generation system, comprising: The data access and logical modeling unit acquires the original business dataset from multiple data source modules, extracts the field attributes and inter-table relationships of each data table in the original business dataset, performs data type standardization processing on the field attributes, and constructs a multi-dimensional logical model based on the inter-table relationships; The semantic mapping and indicator configuration unit extracts the physical name and data type of each field in the multidimensional logical model, constructs a virtual semantic layer, translates each field into the corresponding business semantic entity through the virtual semantic layer, generates a dimension set and a measurement set, calculates measurement rules and binds each measurement item in the measurement set, and outputs a multidimensional analysis semantic configuration scheme. The multidimensional filtering and real-time query unit obtains the multidimensional filtering condition group and graphic display type input by the report user, performs filtering matching and multidimensional association combination on the dimension set and metric set in the multidimensional analysis semantic configuration scheme based on the multidimensional filtering condition group, performs multidimensional real-time query based on the association combination result, and generates a multidimensional result dataset. The report assembly and visualization publishing unit acquires the multidimensional result dataset, matches the corresponding visualization chart template based on the graphic display type, performs data binding and layout filling on the multidimensional result dataset and the visualization chart template, and generates a multidimensional analysis report.

[0026] Specifically, the data type standardization process includes: Establish a cross-data source data type mapping benchmark, match the original data type identifiers of each field in different data sources with the data type mapping benchmark, and generate a standardized data type mapping table; Perform format compliance validation on the original data values ​​of each field, and correct any abnormal data values ​​that fail the validation according to the preset abnormal data processing rules; The corrected data values ​​are formatted according to the corresponding standardized data type identifier to generate standardized field data.

[0027] Specifically, the construction process of the multidimensional logical model includes: Extract the primary key and foreign key identifiers of each data table. Based on the matching relationship between the primary key and foreign key identifiers, generate a central fact table. Then, identify the data tables referenced by the central fact table through foreign key associations as the corresponding fact dimension tables. The central fact table is used as the central node of the star-shaped structure of the multidimensional logical model, and each fact dimension table is linked to the peripheral nodes of the central fact table through foreign key associations. Perform fact-dimensional reference validation on the completed multidimensional logical model. After the validation passes, output the final multidimensional logical model.

[0028] Specifically, the construction process of the virtual semantic layer includes: Extract the physical name and data type of each field in the multidimensional logical model, match the physical name with a preset business term dictionary, and generate business semantic entities corresponding to each field; Write the translation mapping relationship between physical fields and business semantic entities into the mapping structure of the virtual semantic layer to establish a bidirectional translation link between the physical layer and the semantic layer. When multiple candidate semantic entity identifiers exist for the same physical field, semantic disambiguation is performed. A unique translation mapping relationship is determined based on the business context relevance of the semantic entity identifier, and a virtual semantic layer is constructed.

[0029] Specifically, the multidimensional analysis semantic configuration scheme includes: Through the bidirectional translation link of the virtual semantic layer, the physical fields in the multidimensional logical model are translated into corresponding business semantic entities; Entities with discrete attributes in the business semantic entity are classified as dimensions, generating a dimension set; entities with continuous numerical values ​​in the business semantic entity are classified as measures, generating a measure set. A preset dynamic binding mechanism for measurement rules is invoked for each measurement item in the measurement set. The measurement rule binding is performed on the measurement item through the dynamic binding mechanism, and a multidimensional analysis semantic configuration scheme is output.

[0030] Specifically, the dynamic binding mechanism for measurement rules includes: Perform a multi-dimensional cross-combination of the dimensions in the dimension set and the measures in the measure set to generate a candidate set of measure items; The metrics that meet the preset dimension level derivation conditions in the metric candidate set are marked as derived metrics, and calculation metric rules are generated based on the dimension combination and aggregation method corresponding to the derived metrics. The generated computational metric rules are bound to the corresponding derived metric items, and default aggregation computational rules are bound to each basic metric item to generate a complete set of metric rule bindings.

[0031] Example 1: This example uses a sales analysis scenario of an auto parts manufacturing company. The company has deployed a dynamically configurable BI multidimensional analysis report generation system, which needs to integrate sales order data from the ERP system, customer data from the CRM system, and payment collection data from the financial system to generate multidimensional analysis reports for management use.

[0032] Step 1: Multi-source data access and data type standardization processing The data access and logical modeling unit acquires raw business datasets from the ERP data source module, CRM data source module, and financial data source module. Taking the ERP sales order table, CRM customer information table, and financial receipt record table as examples, fields with the same semantics in different data sources may have inconsistent data types—for example, the "order amount" field in ERP is of type DECIMAL, the "receipt amount" field in the financial system is of type NUMERIC, the "customer number" field in CRM is of type VARCHAR, while in ERP it is of type CHAR.

[0033] The data access and logical modeling unit establishes a data type mapping benchmark across data sources. This benchmark defines the set of standard data types supported by the system and their corresponding precision and format specifications: the numeric mapping benchmark is NUMERIC, the character mapping benchmark is VARCHAR, and the date mapping benchmark is DATE. The correspondence between the original data type identifiers of each data source and the mapping benchmark is based on the standard data type specifications of the database management system.

[0034] The unit matches the original data type identifiers of each field in different data sources with the data type mapping benchmark: in ERP, DECIMAL type is matched to NUMERIC; in finance, NUMERIC type is directly matched to NUMERIC; in CRM, VARCHAR type is directly matched to VARCHAR; and in ERP, CHAR type is matched to VARCHAR. After matching, a standardized data type mapping table is generated.

[0035] Perform format compliance validation on the raw data values ​​of each field. The validation rules are defined based on the format constraints of each standardized data type: numeric fields are validated for non-numeric characters, date fields for conforming to standard date formats, and character fields for exceeding the maximum length limit. Abnormal data values ​​that fail validation are corrected according to preset abnormal data handling rules: numeric abnormal values ​​are replaced with null values ​​and marked as abnormal; date abnormal values ​​are parsed using multiple formats and the best match is selected; and character values ​​exceeding the length limit are truncated and marked with a truncation flag. The corrected data values ​​are then format-converted according to their corresponding standardized data type identifiers to generate standardized field data.

[0036] Step 2: Construction of a Multidimensional Logical Model The data access and logical modeling unit extracts the field attributes and inter-table relationships from each data table. Taking three data tables as an example: the sales order table contains the primary key "order identifier", the foreign key "customer identifier", the measure fields "order amount" and "order quantity", and the dimension foreign keys "date identifier" and "product identifier"; the customer information table contains the primary key "customer identifier" and the attribute fields "customer name", "region", and "customer level"; the financial receipts table contains the primary key "receipt identifier", the foreign key "order identifier", and the measure field "receipt amount".

[0037] The unit extracts the primary key and foreign key identifiers from each data table, and determines the fact tables and dimension tables in the multidimensional logical model based on the matching relationship between the primary key and foreign key identifiers. The primary key "Order Identifier" of the Sales Orders table is referenced by the foreign key "Order Identifier" of the Financial Receipts table, and contains the numeric measure fields "Order Amount" and "Order Quantity," therefore it is identified as the central fact table. The core characteristics of a fact table are: it contains numeric measure columns, and it links multiple dimension tables through foreign keys. The primary key "Customer Identifier" of the Customer Information table is referenced by the foreign key "Customer Identifier" of the Sales Orders table, and it does not contain numeric measure fields, therefore it is identified as a fact dimension table.

[0038] The central fact table (sales order table) serves as the central node in the star schema of the multidimensional logical model. Each fact dimension table (customer information table, date dimension table, product dimension table, etc.) is linked to the peripheral nodes of the central fact table through its primary key and corresponding foreign key in the fact table. In the star schema, fact tables and dimension tables are related through primary and foreign keys; dimension tables themselves are not linked. The entire schema presents a radial pattern radiating from the center. After the linking is complete, a fact-dimensional reference validation is performed on the multidimensional logical model—checking whether each foreign key association has a corresponding dimension primary key record. Once the validation passes, the final multidimensional logical model is output.

[0039] Step 3: Construction of the Virtual Semantic Layer The semantic mapping and indicator configuration unit extracts the physical names and data types of each field in the multidimensional logical model. Taking the sales order table as an example, the physical field names are technical names such as "order_amt", "order_qty", "cust_id", and "order_date". The unit matches these physical names with a preset business terminology dictionary. The business terminology dictionary is established based on the enterprise business terminology table and data dictionary, which records the correspondence between physical field names and business semantics—for example, "order_amt" corresponds to "order amount", "order_qty" corresponds to "order quantity", and "cust_id" corresponds to "customer number".

[0040] After matching is completed, business semantic entities corresponding to each field are generated. The translation mapping relationship between physical fields and business semantic entities is written into the mapping structure of the virtual semantic layer, establishing a bidirectional translation link between the physical layer and the semantic layer—the corresponding business semantic entity can be queried from the physical field through the translation link, and the source of the corresponding physical field can also be queried from the business semantic entity.

[0041] When multiple candidate semantic entity identifiers exist for the same physical field—for example, the "revenue" field might correspond to either "operating revenue" or "main business revenue"—the unit performs semantic disambiguation. Disambiguation is based on the business context relevance of the semantic entity identifiers: the table name, related fields, and other fields in the same table containing the physical field are extracted, and a context relevance coefficient is calculated between these and each candidate semantic entity identifier. The candidate entity with the highest relevance coefficient is then identified as the unique translation mapping. A virtual semantic layer is constructed after semantic disambiguation.

[0042] Step 4: Generation of Multidimensional Analysis Semantic Configuration Scheme Through a bidirectional translation link in the virtual semantic layer, physical fields in the multidimensional logical model are translated into corresponding business semantic entities. Entities with discrete attributes within the business semantic entities are categorized as dimensions—"Customer Name," "Region," "Customer Level," "Order Date," and "Product Name," etc., are all discrete attribute values, generating a set of dimensions. Entities with continuous numerical values ​​within the business semantic entities are categorized as measures—"Order Amount," "Order Quantity," and "Payment Amount," etc., are all continuous numerical values, generating a set of measures.

[0043] A dynamic binding mechanism for measurement rules is triggered for each measurement item in the measurement set. First, multi-dimensional cross-combinations are performed between the dimensions in the dimension set and the measures in the measurement set: the "Region" dimension is cross-combined with the "Order Amount" and "Order Quantity" measures to generate candidate measures "Order Amount in Each Region" and "Order Quantity in Each Region"; the "Customer Level" dimension is cross-combined with the "Order Amount" measure to generate candidate measures "Order Amount for Customers of Each Level"; and the "Order Date" dimension is cross-combined with the "Order Amount" and "Order Quantity" measures to generate candidate measures "Order Amount in Each Time Period" and "Order Quantity in Each Time Period". All cross-combination results are summarized into a candidate measure set.

[0044] Measures that meet the preset dimension-level derivation conditions in the candidate measure set are marked as derived measures. Dimension-level derivation conditions are defined based on the hierarchical structure of dimension attributes: if a dimension has a hierarchical relationship (e.g., "order date" can drill down to "year → quarter → month"), then the cross-combination of that dimension and measure is marked as a derived measure. Calculation rules are generated based on the dimension combinations and aggregation methods corresponding to the derived measures—aggregation calculation rules are generated for "order amount in each region" (aggregation method is summation), and derivation calculation rules are generated for "order amount in each time period" (summing at the "year" level and then performing mean derivation at the "quarter" level). Calculated measures are virtual field values ​​defined by multidimensional expressions; they exhibit different aggregation results depending on various filters and dimensions without changing the source data. The generated calculated measure rules are bound to the corresponding derived measures, outputting a multidimensional analysis semantic configuration scheme.

[0045] Specifically, the combination of filtering and matching with multidimensional association includes: The filter condition elements corresponding to each filter condition in the multidimensional filter condition group are parsed, and the parsed filter condition elements are dynamically mapped and matched with the dimension set and metric set in the multidimensional analysis semantic configuration scheme to generate a filter matching mapping table. Extract the logical operators between each filtering condition in the filtering matching mapping table, and combine the matched dimension field and metric field into a multidimensional association structure based on the logical operators to generate a dimension-metric association graph. The filtering condition elements include the target dimension field, comparison operators, and condition values.

[0046] Specifically, the dynamic mapping matching includes: Extract the target dimension field from the filtering condition elements, perform semantic similarity matching between the target dimension field and the business semantic entity in the virtual semantic layer, and determine the corresponding dimension field. Extract the condition values ​​and comparison operators from the filtering condition elements, verify the legality of the condition values ​​and comparison operators based on the semantic range of the dimension field, and establish a filtering binding relationship between the verified condition values ​​and the corresponding business semantic entities in the virtual semantic layer. For condition values ​​that fail validation, semantic re-parsing is triggered. Based on the re-parsing result, the binding relationship is updated, and a complete filtering and matching mapping table is generated.

[0047] Specifically, the generation process of the dimension-metric association graph includes: Based on the logical operators between the filtering conditions in the filtering matching mapping table, the matched dimension fields and metric fields are associated according to the connection rules corresponding to the logical operators to generate an initial association structure. The association strength is calculated for the connection edges between dimension nodes and metric nodes in the initial association structure. The connection edges are then filtered based on the association strength to form a set of dimension-metric association edges. The association strength is determined based on the association frequency of dimensions and metrics in historical queries. By performing connectivity checks on the dimension-metric association edge set, nodes with connected paths are combined as connected subgraphs to generate the dimension-metric association graph.

[0048] Specifically, the multidimensional real-time query includes: The dimension-metric association graph is deconstructed into a set of dimension nodes and a set of metric nodes. The computational metric rule references of each metric node in the metric node set are extracted, and a parsing binding between the metric node and the computational metric rule is established. Based on the hierarchical relationship of the dimension nodes in the dimension node set and the inter-table relationship in the multidimensional logical model, semantic reconstruction is performed on the parsed and bound metric nodes to generate a multidimensional query intermediate structure. The metric reference expansion operation is performed on the metric nodes that reference the metric calculation rules in the multidimensional query intermediate structure. The expanded metric references are then assembled with the corresponding aggregation operations into a multidimensional aggregation operation queue to generate a multidimensional result dataset.

[0049] Example 2: This example follows from Example 1. Sales personnel need to dynamically filter the "quarterly order amount trends of customers at various levels in the East China region" and display them as a trend chart.

[0050] Step 1: Parsing and Dynamic Mapping Matching of Filtering Conditions The report user inputs a multi-dimensional filter condition group: the dimension filter condition is "Region = East China" and "Order Date = This Year", and the graph display type is "Line Chart".

[0051] The multi-dimensional filtering and real-time query unit parses the filtering elements corresponding to each filtering condition, including the target dimension field, comparison operator, and condition value. It performs semantic similarity matching between the parsed target dimension field and the business semantic entity in the virtual semantic layer to determine the corresponding dimension field. It extracts the condition value and comparison operator, and verifies their validity based on the semantic range of the dimension field—querying the set of valid enumeration values ​​for the "Region" dimension and verifying whether "East China" is within the valid set. If the verification passes, it establishes a filtering binding relationship between the condition value and the corresponding business semantic entity in the semantic layer. For condition values ​​that fail verification, it triggers semantic re-parsing, calculates the similarity between the condition value and each value in the set of valid enumeration values, selects the enumeration value with the highest similarity as the re-parsing result, updates the binding relationship, and generates a complete filtering matching mapping table.

[0052] Step 2: Generation of Dimension-Measure Relationship Graph Extract the logical operators between the filter conditions in the filter matching mapping table. In this embodiment, "Region = East China" and "Order Date = This Year" have an "AND" logical relationship. Based on the concatenation rules corresponding to the "AND" logical operator, associate the dimension fields "Region", "Order Date", and "Customer Level" with the measure field "Order Amount" to generate the initial association structure.

[0053] The association strength is calculated for the edges connecting dimension nodes and metric nodes in the initial association structure. Association strength is determined based on the statistical correlation coefficient between the dimension and metric—calculated based on the frequency with which the dimension is referenced in historical queries for that metric; higher frequency results in a stronger association strength. Edges are then filtered based on association strength, prioritizing the "Order Date" and "Customer Level" dimension nodes, which have the strongest association with "Order Amount," forming a set of dimension-metric association edges. Connectivity is validated on this set; if valid, a dimension-metric association graph is generated, where dimension nodes and metric nodes form a complete connected structure through the AND logic of the filtering conditions.

[0054] Step 3: Execute multidimensional real-time query The dimension-metric relationship graph is deconstructed into a set of dimension nodes and a set of metric nodes. The computational metric rule references for each metric node in the metric node set are extracted, and a parsing binding is established between the metric node and the computational metric rule. Based on the hierarchical relationship (year → quarter → month) of the "order date" dimension and the foreign key relationship between the fact table and dimension table in the multidimensional logical model, semantic reconstruction is performed on the parsed and bound metric nodes to generate a multidimensional query intermediate structure.

[0055] The metric reference expansion operation is performed on the metric nodes that reference the calculated metric rules in the multidimensional query intermediate structure. "Order Amount" is bound to derived calculation rules executed at the "Year → Quarter" level. The expansion operation replaces the "Year-level Order Amount" and "Quarter-level Order Amount" referenced in this rule with their corresponding aggregation expressions. The expanded metric references and their corresponding aggregation operations are assembled into a multidimensional aggregation operation queue according to their dependencies: first, the year-level aggregation operation is executed, and then the quarter-level aggregation operation is executed based on the year-level result. The queue is then executed to generate a multidimensional result dataset.

[0056] Specifically, the data binding and layout filling include: Perform hierarchical parsing on the preset data binding points of each chart component in the visualization chart template to determine the dimension level and metric aggregation granularity corresponding to each data binding point; The data in the multidimensional result dataset are grouped and adapted according to the dimension hierarchy and metric aggregation granularity. The adapted data groups are then mapped to the corresponding data binding points for hierarchical data binding. After data binding, each chart component is adaptively filled according to the layout structure of the visualization chart template. After filling, the inter-component linkage relationship is established for each chart component, and a multi-dimensional analysis report is generated.

[0057] Specifically, the data grouping adaptation includes: Extract the required dimensional hierarchy and metric aggregation granularity for each data binding point, and match the aggregation granularity of the multidimensional result dataset with the required granularity of the data binding points; For data whose aggregation granularity does not match the granularity required by the data binding point, perform granularity re-aggregation operation to aggregate the data to the granularity level required by the data binding point; Data is grouped layer by layer from coarse-grained to fine-grained according to the dimensional hierarchy, and a hierarchical mapping and binding relationship is established between each group of data and the corresponding data binding point at the granularity level.

[0058] Example 3: After the multidimensional real-time query is completed, the system needs to render the multidimensional result dataset into a visual line chart report and publish it.

[0059] Step 1: Visual Chart Template Matching and Hierarchical Analysis The report assembly and visualization publishing unit extracts the graphic display type identifier and matches it with the template type identifier in the chart template library, loading the corresponding visualization line chart template. It then performs hierarchical parsing on the preset data binding points of each chart component in the template to determine the dimension level and metric aggregation granularity corresponding to each data binding point—the horizontal axis component requires a dimension level of "time dimension—quarter level," and the legend grouping component requires a dimension level of "customer level."

[0060] Step 2: Data Grouping and Adaptation and Granular Re-aggregation Perform data grouping and adaptation on the data in the multidimensional results dataset according to the dimensional hierarchy and metric aggregation granularity. Extract the required dimensional hierarchy and metric aggregation granularity for each data binding point, and match the aggregation granularity of the multidimensional results dataset with the required granularity of the data binding points.

[0061] When the chart template requires the vertical axis data binding points to display "monthly" granularity, the quarterly aggregation granularity of the multidimensional results dataset does not match the required granularity of the data binding points. To resolve this mismatch, a granularity re-aggregation operation is performed—calling the date dimension table in the multidimensional logical model to obtain the mapping relationship between quarters and months, and then allocating the quarterly aggregated values ​​to the monthly level according to the historical weight coefficients of the months included in that quarter. Granularity re-aggregation can dynamically resolve data granularity conflicts, allowing the same dataset to adaptively adapt to visualization components with different granularity requirements.

[0062] After re-aggregation, data grouping is performed layer by layer from coarse-grained to fine-grained according to the dimension hierarchy. A hierarchical mapping and binding relationship is established between each group of data and the corresponding data binding point at the granularity level. The adapted data groups are then mapped to the corresponding data binding points, and hierarchical data binding is performed.

[0063] Step 3: Adaptive Layout Population and Report Publishing After data binding, each chart component undergoes adaptive layout filling according to the layout structure of the visualization chart template. After filling, inter-component linkages are established—the time interval selector of the horizontal axis component establishes a linked filtering relationship with the data label component, and the customer level selector of the legend component establishes a linked highlighting relationship with the data label component. These linkages allow the report to automatically adjust the data filling granularity and layout structure when filter conditions change, enabling dynamically configurable generation of multidimensional analysis reports.

[0064] After the report is generated, the report identifier and the output format requirements of the target publishing end are extracted, and multi-end format adaptive conversion is performed. The converted publishing format report file is pushed to the report output end, the publishing status receipt is received, and the report's publishing status flag is updated, completing the closed-loop update of the publishing status.

[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A dynamically configurable BI multidimensional analysis report generation system, characterized in that, include: The data access and logical modeling unit acquires the original business dataset from multiple data source modules, extracts the field attributes and inter-table relationships of each data table in the original business dataset, performs data type standardization processing on the field attributes, and constructs a multi-dimensional logical model based on the inter-table relationships; The semantic mapping and indicator configuration unit extracts the physical name and data type of each field in the multidimensional logical model, constructs a virtual semantic layer, translates each field into the corresponding business semantic entity through the virtual semantic layer, generates a dimension set and a measurement set, calculates measurement rules and binds each measurement item in the measurement set, and outputs a multidimensional analysis semantic configuration scheme. The multidimensional filtering and real-time query unit obtains the multidimensional filtering condition group and graphic display type input by the report user, performs filtering matching and multidimensional association combination on the dimension set and metric set in the multidimensional analysis semantic configuration scheme based on the multidimensional filtering condition group, performs multidimensional real-time query based on the association combination result, and generates a multidimensional result dataset. The report assembly and visualization publishing unit acquires the multidimensional result dataset, matches the corresponding visualization chart template based on the graphic display type, performs data binding and layout filling on the multidimensional result dataset and the visualization chart template, and generates a multidimensional analysis report.

2. The system according to claim 1, characterized in that, The data type standardization process specifically includes: Establish a cross-data source data type mapping benchmark, match the original data type identifiers of each field in different data sources with the data type mapping benchmark, and generate a standardized data type mapping table; Perform format compliance validation on the original data values ​​of each field, and correct any abnormal data values ​​that fail the validation according to the preset abnormal data processing rules; The corrected data values ​​are formatted according to the corresponding standardized data type identifier to generate standardized field data.

3. The system according to claim 1, characterized in that, The construction process of the multidimensional logical model specifically includes: Extract the primary key and foreign key identifiers of each data table. Based on the matching relationship between the primary key and foreign key identifiers, generate a central fact table. Then, identify the data tables referenced by the central fact table through foreign key associations as the corresponding fact dimension tables. The central fact table is used as the central node of the star schema in the multidimensional logical model, and each fact dimension table is linked to the peripheral nodes of the central fact table through foreign key associations. Perform fact-dimensional reference validation on the completed multidimensional logical model. After the validation passes, output the final multidimensional logical model.

4. The system according to claim 1, characterized in that, The construction process of the virtual semantic layer specifically includes: Extract the physical name and data type of each field in the multidimensional logical model, match the physical name with a preset business term dictionary, and generate business semantic entities corresponding to each field; Write the translation mapping relationship between physical fields and business semantic entities into the mapping structure of the virtual semantic layer to establish a bidirectional translation link between the physical layer and the semantic layer. When multiple candidate semantic entity identifiers exist for the same physical field, semantic disambiguation is performed. A unique translation mapping relationship is determined based on the business context relevance of the semantic entity identifier, and a virtual semantic layer is constructed.

5. The system according to claim 1, characterized in that, The multidimensional analysis semantic configuration scheme specifically includes: Through the bidirectional translation link of the virtual semantic layer, the physical fields in the multidimensional logical model are translated into corresponding business semantic entities; Entities with discrete attributes in the business semantic entity are classified as dimensions, generating a dimension set; entities with continuous numerical values ​​in the business semantic entity are classified as measures, generating a measure set. A preset dynamic binding mechanism for measurement rules is invoked for each measurement item in the measurement set. The measurement rule binding is performed on the measurement item through the dynamic binding mechanism, and a multidimensional analysis semantic configuration scheme is output.

6. The system according to claim 5, characterized in that, The dynamic binding mechanism for measurement rules specifically includes: Perform a multi-dimensional cross-combination of the dimensions in the dimension set and the measures in the measure set to generate a candidate set of measure items; The metrics that meet the preset dimension level derivation conditions in the metric candidate set are marked as derived metrics, and calculation metric rules are generated based on the dimension combination and aggregation method corresponding to the derived metrics. The generated computational metric rules are bound to the corresponding derived metric items to generate a complete set of metric rule bindings.

7. The system according to claim 1, characterized in that, The combination of filtering and matching with multidimensional association specifically includes: The filter condition elements corresponding to each filter condition in the multidimensional filter condition group are parsed, and the parsed filter condition elements are dynamically mapped and matched with the dimension set and metric set in the multidimensional analysis semantic configuration scheme to generate a filter matching mapping table. Extract the logical operators between the filtering conditions in the filtering matching mapping table, and combine the matched dimension fields and metric fields into a multi-dimensional association structure based on the logical operators to generate a dimension-metric association graph.

8. The system according to claim 7, characterized in that, The dynamic mapping matching specifically includes: Extract the target dimension field from the filtering condition elements, perform semantic similarity matching between the target dimension field and the business semantic entity in the virtual semantic layer, and determine the corresponding dimension field. Extract the condition values ​​and comparison operators from the filtering condition elements, verify the legality of the condition values ​​and comparison operators based on the semantic range of the dimension field, and establish a filtering binding relationship between the verified condition values ​​and the corresponding business semantic entities in the virtual semantic layer. For condition values ​​that fail validation, semantic re-parsing is triggered. Based on the re-parsing result, the binding relationship is updated, and a complete filtering and matching mapping table is generated.

9. The system according to claim 7, characterized in that, The generation process of the dimension-metric association graph specifically includes: Based on the logical operators between the filtering conditions in the filtering and matching mapping table, the matched dimension fields and metric fields are associated according to the connection rules corresponding to the logical operators to generate an initial association structure. Perform association strength calculation on the connection edges between dimension nodes and metric nodes in the initial association structure to form a set of dimension-metric association edges; By performing connectivity checks on the dimension-metric association edge set, nodes with connected paths are combined as connected subgraphs to generate the dimension-metric association graph.

10. The system according to claim 1, characterized in that, The multidimensional real-time query specifically includes: The dimension-metric association graph is deconstructed into a set of dimension nodes and a set of metric nodes. The calculation metric rule references of each metric node in the metric node set are extracted, and a parsing binding between the metric node and the calculation metric rule is established. Based on the hierarchical relationship of the dimension nodes in the dimension node set and the inter-table relationship in the multidimensional logical model, semantic reconstruction is performed on the parsed and bound metric nodes to generate a multidimensional query intermediate structure. The metric reference expansion operation is performed on the metric nodes that reference the metric calculation rules in the multidimensional query intermediate structure. The expanded metric references are then assembled with the corresponding aggregation operations into a multidimensional aggregation operation queue to generate a multidimensional result dataset.

11. The system according to claim 1, characterized in that, The data binding and layout filling specifically include: Perform hierarchical parsing on the preset data binding points of each chart component in the visualization chart template to determine the dimension level and metric aggregation granularity corresponding to each data binding point; The data in the multidimensional result dataset are grouped and adapted according to the dimension hierarchy and metric aggregation granularity. The adapted data groups are then mapped to the corresponding data binding points for hierarchical data binding. After data binding, each chart component is filled with an adaptive layout according to the layout structure of the visualization chart template. After filling, the inter-component linkage relationship is established for each chart component, and a multi-dimensional analysis report is generated.

12. The system according to claim 11, characterized in that, The data packet adaptation specifically includes: Extract the required dimensional hierarchy and metric aggregation granularity for each data binding point, and match the aggregation granularity of the multidimensional result dataset with the required granularity of the data binding points; For data whose aggregation granularity does not match the granularity required by the data binding point, perform granularity re-aggregation operation to aggregate the data to the granularity level required by the data binding point; Data is grouped layer by layer from coarse-grained to fine-grained according to the dimensional hierarchy, and a hierarchical mapping and binding relationship is established between each group of data and the corresponding data binding point at the granularity level.

Images