Time-series Analysis

Analysis of time series includes a large number research topics. In this field my research activity addresses issues related to:

1. Anomaly and novelty detection;
2. Interpretable clustering of multi-variate time series;
3. Forecasting in irregularly sample time series.

Time2Feat for interpretable clustering of multi-variate time series

Traditional systems prioritize effectiveness, efficiency, and scalability but often lack interpretability, essential for human understanding in critical scenarios.

To tackle this issue, we developed Time2Feat, an innovative machine learning system for clustering multivariate time series (MTS). Time2Feat extracts interpretable features from the time series and uses dimensionality reduction to enhance interpretability by selecting the most informative features. The system also supports semi-supervised learning, allowing domain experts to guide the clustering process with a small amount of labeled data. This approach improves both accuracy and clarity of the results. Experiments on multiple benchmarking datasets demonstrate that Time2Feat outperforms existing methods in effectiveness, interpretability, efficiency, and robustness.

Angela Bonifati, Francesco Del Buono, Francesco Guerra, Miki Lombardi, Donato Tiano: Interpretable Clustering of Multivariate Time Series with Time2Feat. Proc. VLDB Endow. 16(12): 3994-3997 (2023)

Angela Bonifati, Francesco Del Buono, Francesco Guerra, Donato Tiano: Time2Feat: Learning Interpretable Representations for Multivariate Time Series Clustering. Proc. VLDB Endow. 16(2): 193-201 (2022)