User Cycle Mining via FP-Tree to Reduce Memory and I/O
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cyclic pattern mining algorithms in the LBSN field require repetitive scanning of raw data sets, leading to inefficient execution and high memory usage, especially due to their complexity which depends on the size of the given period.
Innovation Solution
A system and method for mining user cyclic patterns that utilizes a user behavior recording module, a time sequence generation module, and a cyclic pattern acquisition module, employing the FP-Growth algorithm to generate cyclic patterns by scanning the data set only twice, reducing I/O operations and memory usage through item position information and the creation of an FP-Tree.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If frequent pattern mining algorithms are used for cyclic pattern mining, then cyclic patterns can be acquired, but the algorithm complexity increases and requires repetitive scanning of raw data sets
Solution Approach 1:
The patent segments the cyclic pattern mining process into distinct phases: data preprocessing, FP-Tree construction, and pattern extraction. By dividing the raw data into transaction records and organizing them into a structured format suitable for FP-Growth algorithm, the complexity is managed through modular processing steps rather than handling raw data directly
Solution Approach 2:
The patent performs preliminary actions by pre-processing the data into a standardized transaction format before applying the mining algorithm. The FP-Tree is constructed in advance from the pre-processed data, allowing the actual pattern mining to operate on a prepared data structure rather than raw data, thus reducing algorithmic complexity
2Reliability
If frequent cyclic patterns are acquired by repetitive scanning of raw data set, then comprehensive cyclic patterns can be found, but execution efficiency decreases and I/O operations increase
Solution Approach 1:
The patent performs preliminary data preprocessing to convert raw data into transaction records with standardized formats, including extracting relevant attributes and organizing them in a suitable structure. This preliminary action ensures that subsequent mining operations work with pre-organized data, eliminating the need for repetitive scanning of raw data sets
Solution Approach 2:
The patent introduces an intermediary data structure (FP-Tree) that mediates between the raw data and the pattern mining process. The FP-Tree compactly represents the pre-processed transaction data, allowing efficient pattern extraction without repeatedly accessing the original data source, thus reducing I/O operations
3Reliability
If frequent cyclic patterns are acquired by repetitive scanning of raw data set, then cyclic patterns can be mined, but memory usage increases
Solution Approach 1:
The patent extracts only the necessary information from raw data during preprocessing, creating compact transaction records that contain only relevant attributes for cyclic pattern mining. By taking out and retaining only essential data elements, the memory footprint is significantly reduced compared to storing or repeatedly loading entire raw data sets
Solution Approach 2:
The patent creates a compact copy of the data in FP-Tree structure that preserves the essential patterns and relationships needed for mining, while using minimal memory space. The FP-Tree is a compressed representation that captures frequency and association information without requiring the full raw data to be held in memory during pattern extraction
Data Source
AI summary
A method for mining user cyclic patterns comprises steps: S1: a user behavior recording module recording time of behaviors of each user; S2: a user behavior time sequence generation module performing batch processing to the behaviors of each user to generate a time sequence of the behaviors of the user by using day as time granularity; and S3: a user behavior cyclic pattern acquisition module generating time sequence cyclic patterns from the time sequence according to frequent patterns.