A Multiplicative Benchmarking Framework for Algorithmic Stability and Explanation Resiliency (FASER) In Time-Series AI

Conventional time-series model evaluation relies on point-estimate error metrics like Mean Absolute Percentage Error (MAPE) on static holdout sets, masking structural fragility and attribution drift under real-world operational stress. This paper introduces the Framework for Algorithmic Stability and Explanation Resiliency (FASER), an axiomatic composite benchmarking protocol that evaluates non-linear time-series architectures across three coupled operational pillars: exponential baseline accuracy (S_Perf), logistic noise degradation resilience under Monte Carlo feature perturbations (S_Rob), and entropy-adjusted feature attribution stability across rolling temporal windows (S_Stab). Rather than using linear additive scoring, FASER implements a multiplicative Cobb-Douglas aggregation structure gated by indicator penalty functions, preventing high clean accuracy from obscuring severe noise sensitivity or explainability drift. Theoretical evaluation across gradient-boosted decision trees, random forests, multi-layer perceptrons, and autoregressive econometric baselines demonstrates that deep architectures suffer severe error propagation under stochastic noise, inflating the Noise Degradation Ratio (NDR) and reducing S_Rob. Tree ensembles maintain structural resilience, yet gradient boosting exhibits local feature importance fluctuations across consecutive windows, lowering the Feature Importance Stability Index (FISI / S_Stab). By enforcing hard performance thresholds and geometric coupling across accuracy, perturbation resilience, and attribution stability, FASER provides a rigorous methodology to eliminate brittle predictive architectures prior to deployment in safety-critical environments.

AI Model Robustness, Time-Series Forecasting, Cobb-Douglas Aggregation

1. M. Sudha et al., "Bayesian-optimized ensemble machine learning with fuzzy boundary conditions for robust subsurface temperature forecasting around high-voltage buried cables," Journal of Thermal Analysis and Calorimetry, 2026, doi: 10.1007/s10973-026-15972-9. [Google Scholar] [Crossref]

2. S.-E. Tsai et al., "Resilience-Oriented Daily Peak Load Forecasting on Utility Open Data Using a Dual-Stream Seasonal ARIMA and Kolmogorov–Arnold Network," IEEE Access, vol. 14, pp. 37606–37618, 2026, doi: 10.1109/ACCESS.2026.3659221. [Google Scholar] [Crossref]

3. I.-T. Yang and H. Prayogo, "Chained output information transfer framework for multi-property concrete prediction incorporating supplementary cementitious materials and recycled aggregate," Journal of Building Engineering, vol. 127, Art. no. 116407, 2026, doi: 10.1016/j.jobe.2026.116407. [Google Scholar] [Crossref]

4. M. Rosa-Morales et al., "Physically Interpretable and AI-Powered Applied-Field Thrust Modelling for Magnetoplasmadynamic Space Thrusters Using Symbolic Regression: Towards More Explainable Predictions," Aerospace, vol. 13, no. 3, Art. no. 245, 2026, doi: 10.3390/aerospace13030245. [Google Scholar] [Crossref]

5. S. Zhang et al., "Integrated machine-learning modelling for mechanical property prediction – A case study on laser-welded TC4 titanium alloy," Optics and Laser Technology, vol. 199, Art. no. 115019, 2026, doi: 10.1016/j.optlastec.2026.115019. [Google Scholar] [Crossref]

6. Lalita et al., "A Unified Discrete Mathematical Framework for Validation and Feature Attribution of Trust in Hybrid AI-Driven Network Architectures," Journal of Intelligent Decision Making and Information Science, vol. 3, no. 7S, pp. 1913–1931, 2026, doi: 10.59543/jidmis.v3.1904. [Google Scholar] [Crossref]

7. F. Kurker, "Reliable Estimation of Neutral Current in Industrial Power Systems Using Genetic Algorithm–Based Ensemble Learning and Multimethod Explainability Analysis," International Transactions on Electrical Energy Systems, vol. 2025, no. 1, Art. no. 9960546, 2025, doi: 10.1155/etep/9960546. [Google Scholar] [Crossref]

8. M. Cui et al., "Seismic Deformation Capacity Prediction of Steel-Reinforced Concrete (SRC) Columns Based on Test Database and Machine Learning," Buildings, vol. 16, no. 10, Art. no. 1891, 2026, doi: 10.3390/buildings16101891. [Google Scholar] [Crossref]

9. Q. Wang et al., "Machine learning-driven feature importance appraisal of seismic parameters on tunnel damage and seismic fragility prediction," Engineering Applications of Artificial Intelligence, vol. 137, Art. no. 109101, 2024, doi: 10.1016/j.engappai.2024.109101. [Google Scholar] [Crossref]

10. A. Gorji Sefidmazgi et al., "Resilience and Recovery of Points of Interest Following Hurricane Helene: A Temporal Analysis," Transportation Research Record, 2026, doi: 10.1177/03611981261455026. [Google Scholar] [Crossref]

11. S. Liu and Z. Wang, "Real-time effluent water quality prediction model based on BiLSTM and KAN for wastewater treatment plants," Journal of Water Process Engineering, vol. 78, Art. no. 108750, 2025, doi: 10.1016/j.jwpe.2025.108750. [Google Scholar] [Crossref]

12. W. Wang et al., "Interpretable Attribution of Sentinel-1/2 and Environmental Covariates for Compositionally Closed Soil Mapping and Uncertainty Quantification," Remote Sensing, vol. 18, no. 12, Art. no. 2051, 2026, doi: 10.3390/rs18122051. [Google Scholar] [Crossref]

• The Role of Artificial Intelligence in Revolutionizing Library Services in Nairobi: Ethical Implications and Future Trends in User Interaction
• ESPYREAL: A Mobile Based Multi-Currency Identifier for Visually Impaired Individuals Using Convolutional Neural Network
• Comparative Analysis of AI-Driven IoT-Based Smart Agriculture Platforms with Blockchain-Enabled Marketplaces
• AI-Based Dish Recommender System for Reducing Fruit Waste through Spoilage Detection and Ripeness Assessment
• SEA-TALK: An AI-Powered Voice Translator and Southeast Asian Dialects Recognition