Gaussian Mixture Model and Independent Component Analysis for Time varying Analysis

Clustering multivariate time series (MTS) by incorporating the related-evolving pattern and identifying common behaviors across various fields like data science, bioinformatics, finance, economics and environmental science are the central role for improving clustering results. Even though clustering time series data has been extensively studied in the past eras, no sufficient consideration has been paid to seizure time-varying association patterns in MTS. We recommend an innovative approach for clustering MTS data based on time-varying parameters. The proposed method introduces an unsupervised learning approach using Gaussian Mixture Model (GMM) to explain the related well-defined model between MTS features and generate time-varying variable identification problem as an optimized result, which enables the appropriate learning algorithm based on correlated block matric coordinates. Additionally, we apply the Independent Component Analysis (ICA) method on high-dimensional sequences to achieve low-dimensional feature vectors, and implement an effective structure for temporal based clustering approach to identify cluster with real world data-set. We incorporate wide experiments to compare the planned approach with 2 different clustering algorithms based on an open MTS dataset, which express that GMM meaningfully outperforms with improved efficiency on a diversity of clustering performance metrics.