What question did this study set out to answer?

This study aims to evaluate the performance of various time series forecasting models on software runtime metrics.

May 16, 2026Journal of Systems and Software0 citationsOpen Access

Comparative Study of Statistical, Neural Network, and Foundation Models for Software Runtime...

Forecasting software runtime metrics: A comparative study of classical statistical, neural network, and foundation models

View Full Paper

Ask AI

Bookmark

Authors

Comparative study evaluates time series forecasting performance in software metrics, suggesting foundation models as effective for quality assurance.

This study aims to evaluate the performance of various time series forecasting models on software runtime metrics.
Conducted a comprehensive empirical evaluation using 110 real-world software runtime metrics over the course of one year.
Assessed three classical statistical models, three neural network models, and two foundation models.
Analyzed model performance in terms of capability to forecast software behavior and identify anomalies.
Foundation models achieved state-of-the-art performance in forecasting software runtime metrics, outperforming classical and neural network models.
Performance differences were statistically significant, indicating the effectiveness of foundation models in a zero-shot setting.
No universally superior model was found, emphasizing variability across different time series data.

Menna et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

May 16, 2026Journal of Systems and Software0 citationsOpen Access

Forecasting software runtime metrics: A comparative study of classical statistical, neural network, and foundation models

View Full Paper

Ask AI

Bookmark

Authors

Comparative study evaluates time series forecasting performance in software metrics, suggesting foundation models as effective for quality assurance.

This study aims to evaluate the performance of various time series forecasting models on software runtime metrics.
Conducted a comprehensive empirical evaluation using 110 real-world software runtime metrics over the course of one year.
Assessed three classical statistical models, three neural network models, and two foundation models.
Analyzed model performance in terms of capability to forecast software behavior and identify anomalies.
Foundation models achieved state-of-the-art performance in forecasting software runtime metrics, outperforming classical and neural network models.
Performance differences were statistically significant, indicating the effectiveness of foundation models in a zero-shot setting.
No universally superior model was found, emphasizing variability across different time series data.

Menna et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark