Business data analysis and intelligent report generation system and method supporting custom dimension

Abstract

The invention belongs to the technical field of data analysis, and particularly relates to a business data analysis and intelligent report generation system and method supporting user-defined dimensionalities, and the method comprises the steps: connecting various business databases of an enterprise through a data integration layer, and aggregating multi-source business data; the data processing storage layer preprocesses and stores the integrated data; the method comprises the following steps: constructing a metadata management library based on business data blood relationship, index caliber definition and dimension attribute information through a core service layer, analyzing a user-defined dimension selection instruction and index calculation logic, calling corresponding result data, completing anomaly detection and trend prediction through an AI engine, marking abnormal data points and outputting a prediction result. And the application layer receives the result, configures an interactive chart template and finally generates an intelligent report, so that the personalized data analysis and report display requirements of the user are met. According to the method and the device, the problem of inaccurate analysis result in the analysis process of the business data in the prior art can be solved.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, and in particular relates to a system and method for business data analysis and intelligent report generation that supports custom dimensions. Background Technology

[0002] In the current business operations of enterprises, data analysis and report generation are core components supporting decision-making, but existing technical solutions have many problems that urgently need to be solved.

[0003] At the data level, business data from various departments within an enterprise is scattered across different databases, independent systems, and even offline Excel files, resulting in a prominent data silo phenomenon. Statistical values ​​for the same business metric often deviate across different reports, making it difficult for decision-makers to obtain unified and reliable data for reference, severely impacting the accuracy of their decisions.

[0004] At the system and tool level, traditional systems generate reports that are mostly static statistical results based on historical data. They can only reflect the business situation that has already occurred, and cannot dynamically monitor real-time business data or provide early warnings of anomalies. They also cannot provide forward-looking suggestions for decision-making, making it difficult to meet the needs of enterprises for real-time decision support.

[0005] At the level of people and processes, most companies still rely on the experience and intuition of managers for decision-making, and the concept of data-driven decision-making has not been effectively implemented. Due to inconsistent data quality and the complexity of analytical tools, employees lack trust in the data, which further hinders the full realization of the data's value.

[0006] The above problems mainly stem from two reasons: First, technological limitations. The technology selection and architecture design of the early systems were constrained by the technological conditions at the time and could not adapt to the massive, real-time, and diverse data processing needs of modern enterprises. Second, the disconnect between technology and business. Most related projects were led and promoted by the IT department, with the core objective focused on system launch and function implementation, rather than providing solutions to the actual decision-making pain points of the business departments. This resulted in a misalignment between technical solutions and business needs, putting the cart before the horse. Summary of the Invention

[0007] The technical problem solved by this invention is to provide a system and method for business data analysis and intelligent report generation that supports custom dimensions, so as to solve the problem of inaccurate analysis results in the existing technology for business data analysis.

[0008] The basic solution provided by this invention is a business data analysis and intelligent report generation system that supports custom dimensions, comprising a data integration layer, a data processing and storage layer, a core service layer, and an application layer, wherein: The data integration layer is used to connect to various business databases of an enterprise and integrate business data from multiple sources; The data processing and storage layer is used to preprocess the integrated business data to generate preprocessed business data and then store it. The core service layer is used to read the preprocessed business data stored in the database, build a metadata management library based on the lineage of the business data, the definition of indicators, and the dimension attribute information, parse the user-provided custom dimension selection instructions and indicator calculation logic, retrieve the corresponding result data from the metadata management library, and use the AI ​​engine to perform anomaly detection and trend prediction on the result data, mark abnormal data points and generate trend prediction results. The application layer is used to receive result data, abnormal data points, and trend prediction results generated by the core service layer, configure interactive icon templates, and generate intelligent report results.

[0009] Furthermore, the core service layer includes a metadata management module, a dynamic indicator bullet screen generation module, a user interaction judgment module, and an indicator calculation and generation module, wherein: The metadata management module is used to construct a metadata management library based on the lineage of business data, indicator definitions, and dimensional attribute information; the metadata management library is used for adding, modifying, querying, and versioning metadata. The dynamic indicator bullet screen generation module is used to generate a set of indicator bullet screen blocks based on the dimension and indicator association relationships in the metadata management library, using a multidimensional association degree clustering algorithm, and randomly extracts a preset initial number of indicator bullet screen blocks to present them in bullet screen form; the calculation expression for extracting the preset initial number of indicator bullet screen blocks is:

[0010] in, This indicates the initial number of bullet screen blocks. Indicates the total number of dimensions. Indicates the initial screening coefficient; The user interaction judgment module is used to monitor the user's selection operation of indicator bullet screen blocks in real time. If the user selects any indicator bullet screen block, the indicator bullet screen block is added to the selected set, and the prediction update of the remaining indicator bullet screen blocks and the dynamic reduction of the number of indicator bullet screen blocks are triggered. If the inactivity time exceeds the preset time threshold, it is determined that the user is not satisfied with the current indicator bullet screen block, and a new batch of indicator bullet screen blocks is generated. If the user triggers a confirmation operation or the number of indicator bullet screen blocks in the selected set exceeds the preset threshold of the total number of indicator bullet screen blocks in the set, a custom dimension selection instruction is generated for the indicator bullet screen blocks in the final selected set. The indicator calculation and generation module is used to automatically generate indicator calculation methods based on the indicator definition associated with the selected dimensions in the metadata management library, and present them to the user in a visual form for confirmation.

[0011] Furthermore, the prediction update of the remaining indicator bullet screen block in the user interaction judgment module specifically involves: For each indicator bullet comment block in the selected set, extract the dimensional feature vector. ;in These respectively represent the degree of matching between the dimension and the business objective, the degree of correlation between the dimension and the selected dimensions, the usage rate of the dimension in historical analysis, and the completeness of the dimension data; A multi-head attention mechanism is used to calculate the attention weights of each selected dimension feature. The expression is:

[0012] in, This represents the user's query vector. This indicates the similarity between the query vector and the feature vector of a dimension. Indicates the selected set; The attention-weighted feature vector is input into a bidirectional LSTM prediction network, which outputs the predicted probability distribution of the user's potential needs dimension. The expression is:

[0013] in, This indicates the total number of dimensions that were not selected. Filtering probability The dimension is used as a candidate dimension, where, , This indicates the prediction threshold.

[0014] Furthermore, the dynamic reduction operation of the number of indicator bullet comments in the user interaction judgment module is specifically as follows: The number of candidate dimension indicator bullet comments is determined based on sliding window similarity calculation; wherein, the sliding window size The expression is:

[0015] in, Indicates the selected set; Calculate the average similarity between the candidate dimension and the selected dimension. The remaining number of indicator bullet comments is dynamically reduced using the following expression:

[0016] in, The reduction factor is... .

[0017] Furthermore, in the user interaction judgment module, if the detected inactivity time exceeds a preset time threshold, it is determined that the user is dissatisfied with the current indicator bulletin block, triggering the generation of a new batch of indicator bulletin blocks. Specifically: Calculate the user inactivity time threshold based on historical user interaction data. The expression is:

[0018] in, Indicates the weighting coefficient. Indicates the average historical interaction interval of a user. This represents the base threshold.

[0019] Furthermore, the core service layer also includes an AI intelligent analysis module, which is used to call the AI ​​engine to perform anomaly detection and trend prediction on the result data corresponding to the parsed custom dimensions and indicators, wherein: The anomaly detection employs an improved Isolation Forest algorithm, which introduces a custom dimension weight sparsity optimization to optimize anomaly score calculation. The expression is as follows:

[0020] in, This represents the average path length of data point x. Indicates the reference path length. This represents the weight coefficient corresponding to the custom dimension; The trend prediction employs an ARIMA-LSTM combined model with an attention mechanism to generate the final trend prediction result; the expression for the final trend prediction result is as follows:

[0021] in, The fusion coefficient is... This is the ARIMA trend forecast result. This is the trend prediction result from LSTM.

[0022] Furthermore, the data integration layer includes a business database access module, a streaming data access module, and a third-party data access module, wherein: The business database access module is used to connect to a relational database through a driver and to obtain business data by configuring a scheduled synchronization task or a real-time monitoring mechanism. The streaming data access module is used to deploy Kafka as a message queue to receive real-time log streaming data and click streaming data, and to perform data collection and preliminary filtering through Logstash. The third-party data access module is used to call interfaces via RESTful API, configure API keys and access permissions, and pull data from third-party platforms at a preset frequency.

[0023] Furthermore, the data processing and storage layer includes a preprocessing module, a data storage module, and an index generation module, wherein: The preprocessing module is used to clean and integrate business data to generate high-quality data models. The data storage module is used to build a data warehouse, generate a data table structure according to the star schema, and store preprocessed business data; The index generation module is used to perform pre-aggregation calculations on various dimensions and metrics in business data to generate aggregated indexes.

[0024] Furthermore, the application layer includes an API gateway module, a business API module, a web frontend, and a scheduled task module, wherein: The API gateway module provides identity authentication, permission verification, request rate limiting, and request routing functions, and distributes frontend requests to the corresponding business APIs. The business API module provides data query APIs, report management APIs, and AI service APIs. The web front-end is used to integrate UI component libraries and visualization chart libraries to build drag-and-drop analysis interfaces; The scheduled task module is used to provide scheduled tasks for data querying, AI analysis, and report rendering based on the task scheduling engine, and push them to the specified message recipients.

[0025] A method for business data analysis and intelligent report generation that supports custom dimensions, applied to the aforementioned system for business data analysis and intelligent report generation that supports custom dimensions, includes: S1: Connects to various business databases of the enterprise through the data integration layer to integrate business data from multiple sources; S2: The integrated business data is preprocessed through the data processing and storage layer to generate preprocessed business data, which is then stored. S3: Reads preprocessed business data stored in the core service layer, builds a metadata management library based on the lineage of business data, indicator definition, and dimension attribute information, parses user-provided custom dimension selection instructions and indicator calculation logic, retrieves corresponding result data from the metadata management library, performs anomaly detection and trend prediction on the result data through the AI ​​engine, marks abnormal data points and generates trend prediction results; S4: Receives result data, abnormal data points, and trend prediction results generated by the core service layer through the application layer, configures interactive icon templates, and generates intelligent report results.

[0026] The principles and advantages of this invention are as follows: First, the data integration layer breaks down the barriers between multiple data sources, unifying access to business databases, log streams, and third-party data, thus laying a solid data foundation. Second, the data processing and storage layer improves data quality through standardized preprocessing and adopts a layered storage strategy to balance storage efficiency and query performance. The core service layer, centered on a metadata management library, achieves standardized management of dimensions and indicators by recording data lineage, indicator definitions, and dimensional attributes. It also innovatively analyzes user-defined requirements and combines an AI engine to perform anomaly detection (identifying abnormal data fluctuations) and trend prediction on target data. Finally, the application layer visualizes the analysis results through interactive chart templates, forming a complete "data access-processing-analysis-display" technical closed loop, ensuring the flexibility of customized dimension analysis and the intelligence of report generation.

[0027] The advantages are as follows: Compared with traditional business data analysis systems, this solution has significant advantages in three aspects: data integration, demand adaptation, and decision support. First, data integration is more efficient. Through unified access to multi-source data and hierarchical storage design, it solves the problems of data silos and chaotic storage in traditional systems. At the same time, standardized preprocessing ensures data consistency and provides a reliable data foundation for subsequent analysis. Second, demand adaptation is more flexible. Relying on the metadata management library, it supports users to customize dimension selection and indicator calculation logic, breaking the limitations of fixed dimensions and fixed indicators in traditional systems. It can accurately match personalized analysis needs under different business scenarios. Third, decision support is more intelligent. Through the AI ​​engine, it realizes automatic marking of abnormal data and prediction of business trends. Compared with the ability of traditional systems to only generate static historical reports, it can provide early warning of risks and predict opportunities. Moreover, interactive chart templates make the report results more intuitive and easy to understand, helping decision-makers to quickly obtain effective data insights and improve decision-making efficiency and accuracy. Attached Figure Description

[0028] Figure 1 This is a system architecture diagram of an embodiment of the present invention; Figure 2 This is a flowchart of an embodiment of the present invention. Detailed Implementation

[0029] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 As shown: A business data analysis and intelligent report generation system that supports custom dimensions includes a data integration layer, a data processing and storage layer, a core service layer, and an application layer, wherein: The data integration layer is used to interface with various business databases of the enterprise and integrate business data from multiple sources. In this embodiment, the data integration layer includes a business database access module, a streaming data access module, and a third-party data access module, wherein: The business database access module is used to connect to relational databases such as MySQL, PostgreSQL, and Oracle through JDBC / ODBC drivers, and to configure scheduled synchronization tasks or real-time monitoring mechanisms to obtain business data; The streaming data access module is used to deploy Kafka as a message queue to receive real-time log streaming data and click streaming data, and to perform data collection and preliminary filtering through Logstash. The third-party data access module is used to call interfaces via RESTful API, configure API keys and access permissions, and pull data from third-party platforms at a preset frequency; it supports JSON / XML format data parsing.

[0030] The data processing and storage layer is used to preprocess the integrated business data to generate preprocessed business data, and then stores it. In this embodiment, the data processing and storage layer includes a preprocessing module, a data storage module, and an index generation module, wherein: The preprocessing module is used to clean and integrate business data to generate high-quality data models. Specifically, it uses the Airbyte / DBT tool to configure data extraction rules to obtain data from the source system, and completes data cleaning (duplicate removal, missing value completion, and format standardization) and integration through built-in transformation functions to generate high-quality data models that meet the analysis requirements.

[0031] The data storage module is used to build a data warehouse, generate a data table structure according to the star schema, and store preprocessed business data. Specifically, it uses Snowflake / BigQuery to build the data warehouse, designs the data table structure according to the star schema, stores all historical detailed data, and supports petabyte-level data storage and batch computing.

[0032] The index generation module is used to pre-aggregate and calculate various dimensions and metrics in business data to generate aggregated indexes. Specifically, it deploys ClickHouse / Druid as the OLAP engine to pre-aggregate and calculate commonly used analysis dimensions and metrics to generate aggregated indexes, keeping the multidimensional query response time in the millisecond range.

[0033] The core service layer is used to read preprocessed business data from storage, construct a metadata management library based on the data's lineage, indicator definitions, and dimensional attribute information, parse user-provided custom dimension selection instructions and indicator calculation logic, retrieve corresponding result data from the metadata management library, and use an AI engine to perform anomaly detection and trend prediction on the result data, marking abnormal data points and generating trend prediction results. The core service layer includes a metadata management module, a dynamic indicator bulletin generation module, a user interaction judgment module, an indicator calculation and generation module, and an AI intelligent analysis module. The metadata management module is used to construct a metadata management library based on the lineage of business data, indicator definitions, and dimensional attribute information. This metadata management library is used for adding, modifying, querying, and versioning metadata. In this embodiment, the lineage of business data records the data flow path and transformation rules from the source system to the target storage. Indicator definitions include indicator names, calculation logic, data source tables, and fields. Dimensional attribute information includes dimension names, hierarchical relationships, data types, and enumeration value ranges. The generation of the metadata management library provides a foundation for subsequent custom indicator calculations and access control.

[0034] The dynamic indicator bullet comment generation module is used to generate a set of indicator bullet comment blocks based on the dimension and indicator association relationships in the metadata management library, using a multidimensional association degree clustering algorithm. It then randomly extracts a preset initial number of indicator bullet comment blocks and presents them in bullet comment form. Each indicator bullet comment block stores important information including dimension identifiers, a list of associated indicators, and dimension importance weights. The calculation method for the preset initial number of indicator bullet comment blocks is expressed as follows:

[0035] in, This indicates the initial number of bullet screen blocks. Indicates the total number of dimensions. This represents the initial screening coefficient, which ranges from 0.3 to 0.6 and can be adaptively adjusted according to the complexity of the enterprise's business.

[0036] The user interaction judgment module monitors user selection operations on indicator bullet comments in real time. If a user selects any indicator bullet comment, that bullet comment is added to the selected set, and the prediction update of the remaining indicator bullet comments and the dynamic reduction of the number of indicator bullet comments are triggered. If the inactivity time exceeds a preset threshold, it is determined that the user is not satisfied with the current indicator bullet comments, and a new batch of indicator bullet comments is generated. If the user confirms the selection or the number of indicator bullet comments in the selected set exceeds a preset threshold for the total number of indicator bullet comments in the set, a custom dimension selection instruction is generated for the final selected set of indicator bullet comments. The specific steps to trigger the prediction update of the remaining indicator bulletin blocks are as follows: For each indicator bullet comment block in the selected set, extract the dimensional feature vector. ;in These respectively represent the degree of matching between the dimension and the business objective, the degree of correlation between the dimension and the selected dimensions, the usage rate of the dimension in historical analysis, and the completeness of the dimension data; A multi-head attention mechanism is used to calculate the attention weights of each selected dimension feature. The expression is:

[0037] in, This represents a user query vector, generated by analyzing user history records and the current business scenario. Indicates the selected set; The similarity between the query vector and the feature vector of dimension is represented by a scaled dot product:

[0038] in, The dimension of the feature vector.

[0039] The attention-weighted feature vector is input into a bidirectional LSTM prediction network, which outputs the predicted probability distribution of the user's potential needs dimension. The expression is:

[0040] in, This indicates the total number of dimensions that were not selected. Filtering probability The dimension is used as a candidate dimension, where, , This represents the prediction threshold, which can range from 0.5 to 0.8.

[0041] The specific steps for dynamically reducing the number of indicator bullet comments in the user interaction judgment module are as follows: The number of candidate dimension indicator bullet comments is determined based on sliding window similarity calculation; wherein, the sliding window size The expression is:

[0042] in, Indicates the selected set; Calculate the average similarity between the candidate dimension and the selected dimension. The remaining number of indicator bullet comments is dynamically reduced using the following expression:

[0043] in, This is a reduction factor, ranging from 0.5 to 0.8. Ensure that at least one candidate indicator bullet screen block is generated.

[0044] Then repeat the above steps until the user triggers a confirmation selection or the number of indicator bullet comments in the selected set covers a preset threshold of the total number of indicator bullet comments in the set, such as covering more than 80% of the core dimensions in the metadata. The core dimensions are labeled by business experts and stored in the metadata management library. Finally, the indicator bullet comments in the selected set will generate a custom dimension selection instruction.

[0045] The indicator calculation and generation module is used to automatically generate indicator calculation methods based on the indicator definitions associated with the selected dimensions in the metadata management library, and present them to the user in a visual form for confirmation. In this embodiment, the indicator calculation methods are automatically generated based on the indicator definitions associated with the selected dimensions in the metadata management library, including basic indicators and composite indicators. The basic indicators directly call the preset calculation logic in the metadata, such as "sales revenue = total order amount". The composite indicators are generated based on the user-selected dimensions through a custom formula parsing engine, supporting arithmetic operations, aggregate functions, and conditional judgments, such as "regional gross profit margin = (regional sales revenue - regional cost) / regional sales revenue".

[0046] The AI ​​intelligent analysis module is used to call the AI ​​engine to perform anomaly detection and trend prediction on the parsed result data corresponding to custom dimensions and indicators. The anomaly detection employs an improved Isolation Forest algorithm, which introduces a custom dimension weight sparsity optimization to optimize anomaly score calculation. The expression is as follows:

[0047] in, This represents the average path length of data point x. Indicates the reference path length. This represents the weight coefficient corresponding to the custom dimension; The trend prediction employs an ARIMA-LSTM combined model with an attention mechanism to generate the final trend prediction result; the expression for the final trend prediction result is as follows:

[0048] in, The fusion coefficient is... This is the ARIMA trend forecast result. This is the trend prediction result from LSTM.

[0049] The application layer includes an API gateway module, a business API module, a web frontend, and a scheduled task module, among which: The API gateway module provides identity authentication, permission verification, request rate limiting, and request routing functions, and distributes front-end requests to the corresponding business APIs. In this embodiment, Spring Cloud Gateway is deployed as the API gateway to implement identity authentication (based on JWT tokens), permission verification (integrating with the permission management module), request rate limiting (using the token bucket algorithm and setting single-user / single-interface QPS thresholds), and request routing functions, and distributes front-end requests to the corresponding business APIs.

[0050] The business API module provides data query APIs, report management APIs, and AI service APIs. In this embodiment, the Java / Spring Boot framework is used to develop RESTful APIs such as data query APIs, report management APIs, and AI service APIs. It supports HTTP / HTTPS protocols, provides a unified request parameter format and response format, and supports batch request and asynchronous request processing.

[0051] The web front-end is used to integrate UI component libraries and visualization chart libraries to build a drag-and-drop analysis interface. In this embodiment, the front-end interface is developed based on the React / Vue.js framework, integrating UI component libraries such as Ant Design and visualization chart libraries such as ECharts / Highcharts to build a drag-and-drop analysis interface. Users can select analysis dimensions and indicators through drag-and-drop operations, customize chart types (line charts, bar charts, pie charts, etc.) and report layouts, and support real-time preview of analysis results.

[0052] The scheduled task module is used to provide scheduled tasks for data querying, AI analysis, and report rendering based on the task scheduling engine, and push them to designated message recipients. In this embodiment, Apache Airflow is deployed as the task scheduling engine. Users can configure the report push plan (including execution cycle, push target, and push format) on the front end. Airflow generates scheduled tasks according to the configuration, triggers the data query, AI analysis, and report rendering process, converts the report into PDF / HTML format, and sends it to the designated email address via SMTP protocol, or pushes it to the designated message recipient via the WeChat / DingTalk open platform API.

[0053] In another embodiment of this example, a horizontal management module is also included. This horizontal management module provides support services for access control, monitoring, and alarms at each layer of the system, ensuring the safe and stable operation of the system. Specifically, it includes: Access Control and Auditing: An access control module is designed based on the RBAC (Role-Based Access Control) model. Roles such as administrators, data analysts, and regular users are defined, and different data access permissions (dimension-level and indicator-level access control) and operation permissions (query, edit, export, configuration) are assigned. The system records all user operation logs (including operator, operation time, operation content, and operation result), supporting log querying and audit analysis.

[0054] Monitoring and Alerts: Deploy Prometheus to collect performance metrics (CPU utilization, memory usage, interface response time, data processing throughput) and operational status of each component in the system. Build a monitoring dashboard using Grafana to display the system's operational status in real time. Configure Alertmanager to set alarm thresholds; when metrics exceed the thresholds or components malfunction, send alarm notifications via email, SMS, or messaging tools.

[0055] like Figure 2 As shown, in other embodiments of this example, a method for business data analysis and intelligent report generation supporting custom dimensions is also included, applied to the aforementioned system for business data analysis and intelligent report generation supporting custom dimensions, comprising: S1: Connects to various business databases of the enterprise through the data integration layer to integrate business data from multiple sources; S2: The integrated business data is preprocessed through the data processing and storage layer to generate preprocessed business data, which is then stored. S3: Reads preprocessed business data stored in the core service layer, builds a metadata management library based on the lineage of business data, indicator definition, and dimension attribute information, parses user-provided custom dimension selection instructions and indicator calculation logic, retrieves corresponding result data from the metadata management library, performs anomaly detection and trend prediction on the result data through the AI ​​engine, marks abnormal data points and generates trend prediction results; S4: Receives result data, abnormal data points, and trend prediction results generated by the core service layer through the application layer, configures interactive icon templates, and generates intelligent report results.

[0056] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A business data analysis and intelligent report generation system that supports custom dimensions, characterized in that: It includes a data integration layer, a data processing and storage layer, a core service layer, and an application layer, among which: The data integration layer is used to connect to various business databases of an enterprise and integrate business data from multiple sources; The data processing and storage layer is used to preprocess the integrated business data to generate preprocessed business data and then store it. The core service layer is used to read the preprocessed business data stored in the database, build a metadata management library based on the lineage of the business data, the definition of indicators, and the dimension attribute information, parse the user-provided custom dimension selection instructions and indicator calculation logic, retrieve the corresponding result data from the metadata management library, and use the AI ​​engine to perform anomaly detection and trend prediction on the result data, mark abnormal data points and generate trend prediction results. The application layer is used to receive result data, abnormal data points, and trend prediction results generated by the core service layer, configure interactive icon templates, and generate intelligent report results.

2. The business data analysis and intelligent report generation system supporting custom dimensions according to claim 1, characterized in that: The core service layer includes a metadata management module, a dynamic indicator bullet screen generation module, a user interaction judgment module, and an indicator calculation and generation module, wherein: The metadata management module is used to construct a metadata management library based on the lineage of business data, indicator definitions, and dimensional attribute information; the metadata management library is used for adding, modifying, querying, and versioning metadata. The dynamic indicator bullet screen generation module is used to generate a set of indicator bullet screen blocks based on the dimension and indicator association relationships in the metadata management library, using a multidimensional association degree clustering algorithm, and randomly extracts a preset initial number of indicator bullet screen blocks to present them in bullet screen form; the calculation expression for extracting the preset initial number of indicator bullet screen blocks is: in, This indicates the initial number of bullet screen blocks. Indicates the total number of dimensions. Indicates the initial screening coefficient; The user interaction judgment module is used to monitor the user's selection operation of indicator bullet screen blocks in real time. If the user selects any indicator bullet screen block, the indicator bullet screen block is added to the selected set, and the prediction update of the remaining indicator bullet screen blocks and the dynamic reduction of the number of indicator bullet screen blocks are triggered. If the inactivity time exceeds the preset time threshold, it is determined that the user is not satisfied with the current indicator bullet screen block, and a new batch of indicator bullet screen blocks is generated. If the user triggers a confirmation operation or the number of indicator bullet screen blocks in the selected set exceeds the preset threshold of the total number of indicator bullet screen blocks in the set, a custom dimension selection instruction is generated for the indicator bullet screen blocks in the final selected set. The indicator calculation and generation module is used to automatically generate indicator calculation methods based on the indicator definition associated with the selected dimensions in the metadata management library, and present them to the user in a visual form for confirmation.

3. The business data analysis and intelligent report generation system supporting custom dimensions according to claim 2, characterized in that: The specific steps for triggering the prediction update of the remaining indicator bulletin block in the user interaction judgment module are as follows: For each indicator barrage in the selected set, extract the dimensional feature vector. ;in These respectively represent the degree of matching between the dimension and the business objective, the degree of correlation between the dimension and the selected dimensions, the usage rate of the dimension in historical analysis, and the completeness of the dimension data; A multi-head attention mechanism is used to calculate the attention weights of each selected dimension feature. The expression is: in, This represents the user's query vector. This indicates the similarity between the query vector and the feature vector of a dimension. Indicates the selected set; The attention-weighted feature vector is input into a bidirectional LSTM prediction network, which outputs the predicted probability distribution of the user's potential needs dimension. The expression is: in, This indicates the total number of dimensions that were not selected. Filtering probability The dimension is used as a candidate dimension, where, , This indicates the prediction threshold.

4. The business data analysis and intelligent report generation system supporting custom dimensions according to claim 3, characterized in that: The dynamic reduction operation of the number of indicator bullet comments in the user interaction judgment module is specifically as follows: The number of candidate dimension indicator bullet comments is determined based on sliding window similarity calculation; wherein, the sliding window size The expression is: in, Indicates the selected set; Calculate the average similarity between the candidate dimension and the selected dimension. The remaining number of indicator bullet comments is dynamically reduced using the following expression: in, The reduction factor is... .

5. A business data analysis and intelligent report generation system supporting custom dimensions according to claim 4, characterized in that: In the user interaction judgment module, if the time of no operation pause exceeds a preset time threshold, it is determined that the user is not satisfied with the current indicator bulletin block, and a new batch of indicator bulletin blocks is triggered to be generated. Specifically: Calculate the user inactivity time threshold based on historical user interaction data. The expression is: in, Indicates the weighting coefficient. Indicates the average historical interaction interval of a user. This represents the base threshold.

6. The business data analysis and intelligent report generation system supporting custom dimensions according to claim 2, characterized in that: The core service layer also includes an AI intelligent analysis module, which is used to call the AI ​​engine to perform anomaly detection and trend prediction on the result data corresponding to the parsed custom dimensions and indicators, wherein: The anomaly detection employs an improved Isolation Forest algorithm, which introduces a custom dimension weight sparsity optimization to optimize anomaly score calculation. The expression is as follows: in, This represents the average path length of data point x. Indicates the reference path length. This represents the weight coefficient corresponding to the custom dimension; The trend prediction employs an ARIMA-LSTM combined model with an attention mechanism to generate the final trend prediction result; the expression for the final trend prediction result is as follows: in, The fusion coefficient is... This is the ARIMA trend forecast result. This is the trend prediction result from LSTM.

7. The business data analysis and intelligent report generation system supporting custom dimensions according to claim 1, characterized in that: The data integration layer includes a business database access module, a streaming data access module, and a third-party data access module, wherein: The business database access module is used to connect to a relational database through a driver and to obtain business data by configuring a scheduled synchronization task or a real-time monitoring mechanism. The streaming data access module is used to deploy Kafka as a message queue to receive real-time log streaming data and click streaming data, and to perform data collection and preliminary filtering through Logstash. The third-party data access module is used to call interfaces via RESTful API, configure API keys and access permissions, and pull data from third-party platforms at a preset frequency.

8. The business data analysis and intelligent report generation system supporting custom dimensions according to claim 1, characterized in that: The data processing and storage layer includes a preprocessing module, a data storage module, and an index generation module, wherein: The preprocessing module is used to clean and integrate business data to generate high-quality data models. The data storage module is used to build a data warehouse, generate a data table structure according to the star schema, and store preprocessed business data; The index generation module is used to perform pre-aggregation calculations on various dimensions and metrics in business data to generate aggregated indexes.

9. A business data analysis and intelligent report generation system supporting custom dimensions according to claim 1, characterized in that: The application layer includes an API gateway module, a business API module, a web frontend, and a scheduled task module, wherein: The API gateway module provides identity authentication, permission verification, request rate limiting, and request routing functions, and distributes frontend requests to the corresponding business APIs. The business API module provides data query APIs, report management APIs, and AI service APIs. The web front-end is used to integrate UI component libraries and visualization chart libraries to build drag-and-drop analysis interfaces; The scheduled task module is used to provide scheduled tasks for data querying, AI analysis, and report rendering based on the task scheduling engine, and push them to the specified message recipients.

10. A method for business data analysis and intelligent report generation supporting custom dimensions, applied to a business data analysis and intelligent report generation system supporting custom dimensions as described in any one of claims 1-9, characterized in that: include: S1: Connects to various business databases of the enterprise through the data integration layer to integrate business data from multiple sources; S2: The integrated business data is preprocessed through the data processing and storage layer to generate preprocessed business data, which is then stored. S3: Reads preprocessed business data stored in the core service layer, builds a metadata management library based on the lineage of business data, indicator definition, and dimension attribute information, parses user-provided custom dimension selection instructions and indicator calculation logic, retrieves corresponding result data from the metadata management library, performs anomaly detection and trend prediction on the result data through the AI ​​engine, marks abnormal data points and generates trend prediction results; S4: Receives result data, abnormal data points, and trend prediction results generated by the core service layer through the application layer, configures interactive icon templates, and generates intelligent report results.

Images