A k-MLE framework with its k-VARs algorithm for multivariate time-series clustering

Folders

/ or ROOT: the project root directory.

expt_synthetic/: contains the scripts that simulate data and perform benchmarks that appear in our manuscript.

expt_wafer/: contains the scripts that applies our algorithms on the WAFER dataset, together with the scripts in Python/MATLAB that uses the third-party libraries in the comparative study.

toolkits/: collects the third-party libraries of methods that compared in our manuscript. The scripts directly under this folder are those the author wrote as wrapper functions.

Scripts

Here we list the main scripts/functions that supports the proposed method. For other scripts for specific experiments, refer to Readme.md file in each sub-folder.

• demo.m: a minimal example to apply k-VARs for time-series clustering.

Scripts in functions/ for clustering algorithms and simulator:

• kVARs.m: implements the k-VARs algorithm for clustering.

• simVARs.m: random generation of stable VAR models and simulate to gather a collection of multivariate time series for later clustering tasks.

• calcBIC.m: computes the extended BIC scores to choose model order or the number of clusters, e.g., K, p1,...,pK.

Scripts in measures/ for clustering performance measures:

• perfRI.m: implements Rand Index (RI).

• perfARI.m: implements Adjusted Rand Index (ARI).

• perfNMI.m: implements Normalised Mutual Information (NMI).

• perfNID.m: implements the Normalized Information Distance (NID).

Experiments in our manuscript

• synthetic-data experiments: read Readme.md in the folder expt_synthetic/.

• real-data experiments: read Readme.md in the folder expt_wafer/.

Last modified on 05 Jan 2025.