Association Rule Mining Using Scaled Lift Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data mining techniques face challenges in generating meaningful association rules for complex products due to lack of granularity in item descriptions, leading to redundant rules and increased computational costs, especially when dealing with large and prolific patterns in databases.
Innovation Solution
The system represents items in terms of their features, using part groups and attribute-value pairs to enhance granularity, and employs an adjustable specialization lift threshold to filter out redundant rules, thereby reducing the number of association rules evaluated and improving performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional data mining techniques use generic item descriptions (SKU level) to identify frequent sets, then the item identification process is simple, but the granularity is insufficient to provide meaningful information for complex products
Solution Approach 1:
The patent segments item descriptions into hierarchical levels: generic item level (SKU) and detailed attribute level (feature-name-value triplets). This segmentation allows the system to maintain both simplicity for basic items and granularity for complex products by breaking down product information into constituent features that can be independently analyzed.
Solution Approach 2:
The patent adds a new dimension to item description by introducing attribute-based feature representations alongside traditional SKU identifiers. This dimensional expansion transforms single-dimensional SKU-based identification into multi-dimensional identification incorporating product features, enabling more meaningful pattern discovery while maintaining computational tractability through structured attribute organization.
2Loss of information
If attribute-based association rule mining is used to provide finer granularity, then more meaningful patterns are discovered, but the number of generated rules increases significantly
Solution Approach 1:
The patent implements dynamic rule generation by introducing adjustable confidence thresholds and support thresholds that control the expansion of rule sets. The system dynamically adjusts these parameters based on the complexity of the data and desired outcome, allowing meaningful patterns to be discovered while preventing combinatorial explosion of rules through adaptive threshold management.
Solution Approach 2:
The patent changes key parameters (confidence threshold, support threshold, maximum rule depth) to control the balance between pattern discovery quality and rule set size. By systematically varying these parameters, the system can optimize the trade-off between capturing meaningful attribute-based patterns and maintaining a manageable number of generated rules for practical application.
3Reliability
If multiple passes through large candidate sets are required to identify frequent patterns, then comprehensive pattern discovery is achieved, but computational resources are consumed excessively
Solution Approach 1:
The patent performs preliminary actions by pre-processing transaction data to extract and organize attribute information before the main association rule mining process. Frequent pattern trees are constructed in advance with attribute-based indexing, allowing the system to quickly identify frequent itemsets without requiring multiple exhaustive passes through the entire candidate set, thus reducing computational resource consumption while maintaining discovery completeness.
Data Source
AI summary
A data processing system processes data sets (such as low-resolution transaction data) into high-resolution data sets by mapping generic information into attribute-based specific information that may be processed to identify frequent sets therein. When association rules are generated from such frequent sets, the complexity and/or quantity of such rules may be managed by removing redundancies from the rules, such as by filtering subsumed rules from the generated rule set that have a confidence metric value that does not exceed a first confidence metric value for a subsuming rule by more than a scaled lift threshold value that is calculated by determining a complement of the first confidence metric value, squaring the complement to obtain a squared value and multiplying the squared value by a scaling factor.