Hierarchical Time-Series Clustering with Auto Encoded Compact Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional hierarchical time-series clustering is a time-consuming and computationally inefficient process, especially when dealing with large volumes of data from applications like smart cities and health monitoring, due to the difficulty in finding the right similarity measure and the varying length of time-series data.
Innovation Solution
The method employs Auto Encoded Compact Sequences (AECS) generated using a multilayered Seq2Seq Long Short-Term Memory (LSTM) auto encoder, followed by agglomerative hierarchical clustering with Chebyshev, Mahalanobis, and Manhattan distance measures, and selects the best cluster based on modified Hubert statistics, allowing for efficient clustering of both univariate and multivariate time-series.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional hierarchical time-series clustering is used, then clustering can be performed on time-series data, but the process is highly time-consuming and computationally inefficient
Solution Approach 1:
The patent applies preliminary action by transforming the original time-series data into compact representations (such as shapelets or other condensed forms) before performing hierarchical clustering. This preprocessing step reduces the dimensionality and complexity of the data, enabling faster clustering operations while preserving the essential patterns needed for accurate clustering results.
2Productivity
If conventional hierarchical clustering is applied to time-series data of varying lengths, then clustering can be performed, but the process is not computation efficient
Solution Approach 1:
The patent changes parameters by transforming time-series data of varying lengths into fixed-length compact representations. This parameter transformation standardizes the data format, making it suitable for efficient hierarchical clustering algorithms that require uniform input dimensions, thereby reducing computational complexity.
3Reliability
If conventional approaches are used for time-series clustering, then clustering results can be obtained, but the process is not cost efficient
Solution Approach 1:
The patent extracts the essential patterns and features from the original time-series data into compact representations, separating the critical information from the redundant data. This extraction process maintains clustering accuracy by preserving key patterns while reducing the data volume that requires computational processing, thereby improving cost efficiency.
Data Source
AI summary
Conventional hierarchical time-series clustering is highly time consuming process as time-series are characteristically lengthy. Moreover, finding right similarity measure providing best possible hierarchical cluster is critical to derive accurate inferences from the hierarchical clusters. Method and system for Auto Encoded Compact Sequences (AECS) based hierarchical time-series clustering that enables compact latent representation of time-series using an undercomplete multilayered Seq2Seq LSTM auto encoder followed by generating of HCs using multiple similarity measures is disclosed. Further, provided is a mechanism to select the best HC among the multiple HCs on-the-fly, based on an internal clustering performance measure of Modified Hubert statistic τ. Thus, the method provides time efficient and low computational cost approach for hierarchical clustering for both on univariate and multivariate time-series. AECS approach provides a constant length sequence across diverse length series and hence provides a generalized approach.