Composite Model Object for Partitioned Data Mining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current database systems lack tools to efficiently build, manage, and deploy separate models for capturing behavior across different regions or countries, leading to manageability and performance issues in data mining applications, such as fraud detection and pattern identification.
Innovation Solution
A relational database system that generates a composite model object from partitioned data mining models, allowing for parallelization, reduced memory contention, and dynamic caching, enabling scalable and efficient model building, scoring, and updating of models across multiple processing nodes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If separate customized applications are written to encapsulate separate models for different regions, then model specificity is improved, but manageability and performance deteriorate
Solution Approach 1:
The patent merges multiple separate data mining models into a single composite model object that can handle multiple regions/countries. The system allows building separate models for different geographic regions while managing them through a unified interface, eliminating the need for multiple customized applications. The composite model object consolidates the management of region-specific models, providing a single point of control for deployment and maintenance.
2Adaptability or versatility
If complete customized applications are written to manage separate models, then model functionality is improved, but system complexity increases
Solution Approach 1:
The patent creates a universal composite model object that can serve multiple functions across different regions and use cases. This single object type can encapsulate region-specific models while providing a standardized interface for deployment, scoring, and management. The system handles fraud detection, pattern identification, and predictive analytics through a unified mechanism, reducing the need for region-specific application code.
3Measurement precision
If partitioned models are built for different regions, then pattern identification accuracy is improved, but memory consumption increases
Solution Approach 1:
The patent segments the composite model object into multiple partitions, where each partition contains a region-specific data mining model. This segmentation allows the system to load only the necessary partition into memory when scoring data for a specific region, rather than loading all region-specific models simultaneously. The partitioning strategy enables efficient memory utilization while maintaining the ability to identify region-specific patterns with high accuracy.
Data Source
AI summary
Systems, methods, and other embodiments are disclosed for partitioning models in a database. In one embodiment, a set of training data is parsed into multiple data partitions based on partition keys, where the data partitions are identified by the partition keys and are used for training data mining models. The multiple data partitions are analyzed to generate partition metrics data. Algorithm data, identifying at least one algorithm for processing the multiple data partitions, and resources data, identifying available modeling resources for processing the multiple data partitions, are read. The partition metrics data, the algorithm data, and the resources data are processed to generate an organization data structure. The organization data structure is configured to control distribution and processing of the multiple data partitions across the available modeling resources to generate a composite model object that includes a separately trained data mining model for each partition of the multiple partitions.