Sequence Mining via Resampling and Shuffling Algorithms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional sequence-mining software is inefficient in handling large data sets, consuming excessive computing resources and often failing to identify frequent sequences due to the exponential increase in search space, especially when analyzing sequences of longer lengths.
Innovation Solution
The implementation of a two-step process involving a resampling algorithm to select candidate sequences based on weights and a shuffling algorithm to determine frequencies, followed by a verification step to confirm support levels, reduces computational complexity and resource consumption by sampling directly for k-length sequences without relying on shorter or longer sequence dependencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sequence-mining software is used to analyze large data sets, then the search space is exhaustively explored, but computing resource consumption increases exponentially and execution time becomes unmanageable
Solution Approach 1:
The patent applies partial action by using resampling algorithms to select a representative subset of sequences from the full data set for mining. Instead of exhaustively processing all sequences, the system samples a portion that statistically represents the entire set, thereby reducing computing resource consumption while maintaining acceptable accuracy for identifying frequent sequences.
Solution Approach 2:
The patent changes the parameter of data set size by transforming the original large data set into a resampled subset with controlled size. By adjusting sampling parameters and sequence length thresholds, the system optimizes the balance between identification accuracy and resource consumption, enabling feasible processing of large-scale data.
2Measurement precision
If the search space is exhaustively explored to identify all frequent sequences, then measurement precision is improved, but the complexity of the mining process increases exponentially
Solution Approach 1:
The patent segments the mining process into distinct phases: resampling to generate candidate sequences, frequency calculation on sampled data, and verification against the full data set. This segmentation breaks down the exponentially complex exhaustive search into manageable stages, reducing overall process complexity while maintaining identification accuracy.
Solution Approach 2:
The patent performs preliminary action by pre-calculating frequencies and identifying candidate frequent sequences from the resampled subset before verifying them against the complete data set. This preliminary filtering reduces the number of sequences requiring full verification, thereby simplifying the overall mining process complexity.
3Productivity
If sequences of longer lengths are analyzed, then the value of identified patterns increases, but the search space increases exponentially causing resource exhaustion
Solution Approach 1:
The patent applies partial action by limiting the analysis to sequences up to a maximum length parameter and using resampling to process only a subset of possible sequences. This approach enables the identification of valuable long-length patterns without exhaustively exploring the exponentially growing search space, thus managing resource consumption while maintaining pattern value.
Solution Approach 2:
The patent changes parameters including maximum sequence length and sampling rate to control the growth of search space. By dynamically adjusting these parameters based on available resources and data characteristics, the system maintains the ability to identify high-value long sequences while preventing resource exhaustion.
Data Source
AI summary
Computing resources consumed in performing computerized sequence-mining can be reduced by implementing some examples of the present disclosure. In one example, a system can determine weights for data entries in a data set and then select a group of data entries from the data set based on the weights. Next, the system can determine a group of k-length sequences present in the selected group of data entries by applying a shuffling algorithm. The system can then determine frequencies corresponding to the group of k-length sequences and select candidate sequences from among the group of k-length sequences based on the frequencies thereof. Next, the system can determine support values corresponding to the candidate sequences and then select output sequences from among the candidate sequences based on the support values thereof. The system may then transmit an output signal indicating the selected output sequences an electronic device.