BI System User Interest Clustering via Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Business Intelligence (BI) systems inadequately detect and characterize user interests, leading to tedious and unproductive interactions, as user interests are hidden in interactions and users are often unaware of the data sources and formal queries they trigger.
Innovation Solution
A method that characterizes user interests by extracting features from historical queries, applying similarity measures to group intents into coherent interests, using clustering algorithms like centroid-based, connectivity-based, and density-based clustering, and assigning weights using machine learning models such as support vector machines, to identify and recommend relevant data items based on user interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional query similarity measures are used to group user interactions, then the system can process queries efficiently, but the precision in identifying user interests deteriorates
Solution Approach 1:
The patent transforms the query representation by extracting multiple features (token frequency, refining token frequency, suggestion frequency, BI questions, hierarchical relationships, temporal relationships) and applies weighted similarity measures instead of conventional single-measure approaches. This parameter transformation enables more precise user interest identification while maintaining processing efficiency through structured feature extraction and weighting mechanisms.
2Loss of information
If the system provides detailed data sources and formal queries to users, then users can understand the query process better, but the interaction complexity increases making the system harder to operate
Solution Approach 1:
The patent extracts and separates the complex formal query processing from the user interface by implementing automatic query generation based on natural language input. The system extracts user intent from simple queries, automatically generates formal queries, and presents results without requiring users to understand or interact with the complex underlying data sources and query structures, thus maintaining transparency while preserving ease of operation.
3Measurement precision
If the system collects and analyzes extensive historical query data to identify user interests, then the accuracy of interest detection improves, but the data processing complexity and computational resources increase
Solution Approach 1:
The patent segments the historical query data analysis into distinct feature extraction components (token frequency, refining token frequency, suggestion frequency, BI questions, hierarchical relationships, temporal relationships) that can be processed independently and then combined through weighted similarity measures. This segmentation reduces processing complexity by breaking down the overall analysis task into manageable, modular feature extraction and weighting steps.
Data Source
AI summary
The current subject matter provides systems, methods, computer program products (including graphical user interfaces) in which user interests can be identified by characterizing the intent of the interaction with a data source such as a business intelligence (BI) system. A set of features can be identified for an adequate description of intents, and a similarity measure for grouping intents into coherent interests. Related systems, apparatus, and methods are also disclosed.