![[2602.19984] Multivariate time-series forecasting of ASTRI-Horn monitoring data: A Normal Behavior Model](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.19984] Multivariate time-series forecasting of ASTRI-Horn monitoring data: A Normal Behavior Model
Summary
This article presents a Normal Behavior Model (NBM) for forecasting monitoring data from the ASTRI-Horn telescope, demonstrating effective multivariate time-series predictions.
Why It Matters
The study addresses the need for reliable forecasting in astrophysical instrumentation, enabling early anomaly detection and supporting predictive maintenance systems. This is crucial for enhancing the operational efficiency of telescopes and improving data reliability in astrophysics.
Key Takeaways
- The NBM effectively forecasts multivariate time-series data from ASTRI-Horn.
- MLP model performance matches LSTM while converging faster, achieving low MSE and NMAD.
- The model supports early anomaly detection, crucial for operational monitoring.
- Forecasting horizon can extend to 6.5 hours without performance degradation.
- The study lays groundwork for future prognostics and health management systems.
Astrophysics > Instrumentation and Methods for Astrophysics arXiv:2602.19984 (astro-ph) [Submitted on 23 Feb 2026] Title:Multivariate time-series forecasting of ASTRI-Horn monitoring data: A Normal Behavior Model Authors:Federico Incardona, Alessandro Costa, Farida Farsian, Francesco Franchina, Giuseppe Leto, Emilio Mastriani, Kevin Munari, Giovanni Pareschi, Salvatore Scuderi, Sebastiano Spinello, Gino Tosti View a PDF of the paper titled Multivariate time-series forecasting of ASTRI-Horn monitoring data: A Normal Behavior Model, by Federico Incardona and 9 other authors View PDF Abstract:This study presents a Normal Behavior Model (NBM) developed to forecast monitoring time-series data from the ASTRI-Horn Cherenkov telescope under normal operating conditions. The analysis focused on 15 physical variables acquired by the Telescope Control Unit between September 2022 and July 2024, representing sensor measurements from the Azimuth and Elevation motors. After data cleaning, resampling, feature selection, and correlation analysis, the dataset was segmented into fixed-length intervals, in which the first I samples represented the input sequence provided to the model, while the forecast length, T, indicated the number of future time steps to be predicted. A sliding-window technique was then applied to increase the number of intervals. A Multi-Layer Perceptron (MLP) was trained to perform multivariate forecasting across all features simultaneously. Model performance was evaluat...