Research on load clustering algorithm based on variational autoencoder and hierarchical clustering.

Area of Science:

• Data Science
• Machine Learning
• Artificial Intelligence

Background:

• Time series data clustering faces challenges in feature representation and scalability.
• Existing Variational Autoencoder (VAE) methods struggle with discriminative power and disconnected objectives.

Purpose of the Study:

• To develop a novel deep time series clustering approach integrating VAE with metric learning.
• To improve feature representation, address scalability, and enhance clustering accuracy.

Main Methods:

• Utilized a VAE with Gated Recurrent Units for temporal feature extraction.
• Incorporated metric learning for joint optimization of latent space representation.
• Employed the sum of log likelihoods as the clustering merging criterion.

Main Results:

• Achieved a 27.16% improvement in average clustering accuracy.
• Demonstrated a 47.15% increase in speed on industrial load data.
• Enhanced interpretability of clustering results.

Conclusions:

• The integrated VAE and metric learning approach offers a powerful solution for time series clustering.
• This method provides novel tools for analyzing complex time series data in various domains.
• Further research into VAE applications for time series clustering is warranted.