The researchers set out to evaluate how feature importance and feature selection methods perform when the goal is rapid, reliable estimation of seismic risk levels for reinforced concrete buildings. Rather than treating machine learning as an opaque oracle, they deliberately interrogated it, asking which input parameters drive predictions and whether a trimmed-down set of inputs could match, or even beat, models fed the complete feature set. This dual focus on accuracy and interpretability places the work squarely within the growing field of explainable artificial intelligence, where the goal is not just prediction but understanding.

To build a trustworthy foundation, the team used a simulation-based dataset representing an existing building stock rather than a purely synthetic or idealized collection of structures. This choice matters. Rapid screening methods have a long history in earthquake engineering, from the FEMA P-154 rapid visual screening handbook in the United States to the P25 screening method developed in Türkiye and the Japanese Building Disaster Prevention Association’s evaluation standard. These approaches typically rely on a handful of visually observable parameters. Machine learning promises to go further by capturing subtle, nonlinear interactions among many parameters, but only if the underlying data reflect real buildings with realistic variability.

The study employed three machine learning models to classify seismic risk, allowing the authors to compare how different algorithms weigh and use the same information. Across their analyses, a consistent pattern emerged: combining feature importance measures with explicit feature selection improved prediction performance compared with using either approach alone. Feature importance methods, such as permutation importance, quantify how much each input contributes to a model’s predictions by measuring how performance degrades when that input is shuffled or removed. Feature selection methods, including correlation-based filtering rooted in classical statistical measures such as Pearson’s correlation coefficient, systematically prune redundant or uninformative variables before training.

The standout result came from LightGBM, a gradient boosting framework known for its efficiency on tabular data, when paired with normalized feature importance and correlation-based feature selection. This combination achieved a macro-F1 score of approximately 0.91 — a strong result for a multi-class classification task — while using fewer input parameters than models trained on the full feature set. The macro-F1 metric is particularly meaningful in seismic risk classification because it averages performance across all risk classes, preventing a model from looking good simply by mastering the most common category while neglecting rare but critical high-risk cases.

That a leaner model outperformed the full-feature version is not a paradox but a well-documented phenomenon in machine learning. Irrelevant and redundant features can act as noise, encouraging models to fit spurious patterns that do not generalize. By identifying and retaining only the most influential parameters, the selection process acts as a regularizer, sharpening the model’s attention on the structural characteristics that genuinely govern seismic behavior. For low-rise reinforced concrete buildings, such characteristics typically relate to geometry, structural system, material properties, and code-era design provisions — the same factors that decades of empirical screening methods have converged upon, but now quantified and weighted by data rather than expert judgment alone.

The interpretability dimension is central to the study’s practical value. Engineers and municipal officials cannot act on a black-box score; they need to know why a building is flagged as high risk and which interventions might matter. Frameworks like SHAP, introduced by Lundberg and Lee in 2017, and LIME, proposed by Ribeiro and colleagues in 2016, have made it standard practice to attach explanations to model predictions. By combining normalized feature importance with explicit selection, Yilmaz and Dirikgil provide a transparent chain of reasoning: here are the parameters that matter most, here is the evidence that they suffice, and here is the resulting risk classification. This aligns with a broader movement in structural engineering toward explainable machine learning, exemplified by recent work applying explainable frameworks to XGBoost-based seismic assessment.

The timing of this research is not incidental. The February 2023 Kahramanmaraş earthquake sequence in Türkiye, which included magnitude 7.8 and 7.6 events, caused catastrophic damage to vast numbers of reinforced concrete buildings and prompted intensive post-earthquake damage surveys. Those surveys documented recurring vulnerabilities tied to design era, detailing practices, and construction quality, reinforcing the urgency of pre-earthquake screening. Türkiye’s successive earthquake codes, from the 1975 provisions through the current Turkish Building Earthquake Code of 2018, mean that a building’s design vintage alone carries substantial predictive information. Machine learning models trained on simulation data grounded in these code generations can encode that knowledge and apply it consistently across an entire urban building stock.

What makes the proposed framework genuinely scalable is the economy of its input requirements. Traditional detailed assessment requires structural drawings, material testing, and nonlinear analysis — resources that no city can mobilize for every building. Rapid visual screening reduces the burden but introduces subjectivity and coarse granularity. A model that achieves roughly 0.91 macro-F1 with a reduced set of critical parameters offers a middle path: data-driven consistency at a fraction of the data collection cost. The authors frame this as enabling rapid and efficient large-scale building stock assessments, with direct applications to regional seismic risk evaluation and urban-scale vulnerability screening. For municipalities prioritizing retrofit budgets, the ability to rank thousands of buildings quickly and defensibly could redirect scarce resources to the structures that need them most.

The study also contributes methodologically to the machine learning literature on feature selection. Recent comparative analyses have emphasized that different selection methods vary in both performance and stability, and that the choice of importance measure interacts with the choice of algorithm. By systematically testing combinations of three models with multiple importance and selection strategies, the researchers provide a template that other domains facing high-dimensional structural data can adapt. Their finding that the combination of normalized importance and correlation-based selection with LightGBM yields the best results is an actionable recommendation, not merely an observation. The authors note that some supporting data are included in the article, with expanded datasets available from the corresponding author upon reasonable request, supporting reproducibility in a field where simulation pipelines are often proprietary. As cities worldwide confront aging building stocks and tightening retrofit budgets, the message from Kayseri is clear: predicting which buildings will fail in the next earthquake may require far less information than anyone assumed — provided you know exactly which information to ask for.

Subject of Research: Interpretable machine learning for seismic risk prediction of low-rise reinforced concrete buildings

Article Title: Interpretable machine learning and feature selection for seismic risk prediction of low-rise reinforced concrete buildings

Article References: Yilmaz, M., & Dirikgil, T. (2026). Interpretable machine learning and feature selection for seismic risk prediction of low-rise reinforced concrete buildings. Bulletin of Earthquake Engineering. https://doi.org/10.1007/s10518-026-02667-7

Image Credits: AI Generated

DOI: 10.1007/s10518-026-02667-7

Keywords: seismic risk prediction, machine learning, reinforced concrete buildings, explainable AI, feature selection, feature importance, LightGBM, macro-F1 score, rapid screening, earthquake engineering, urban vulnerability, Turkish building stock

The challenge that motivated the work is structural. Airspace is divided into sectors, and each sector’s complexity depends on two kinds of relationships that are difficult to model together. Spatially, sectors are connected both by simple geographical adjacency, meaning they share a border, and by dynamic traffic flows, since aircraft entering one sector typically exited another, creating dependencies that shift with the daily rhythm of departures and arrivals. Temporally, complexity evolves across multiple time scales at once: fast fluctuations driven by individual climbing and descending aircraft, medium-term waves tied to airport scheduling, and long-range dependencies that link morning traffic management decisions to afternoon congestion. Existing spatio-temporal prediction frameworks, including the graph neural networks that have dominated traffic forecasting in recent years, tend to rely on static graphs that freeze these relationships in place and on shallow temporal alignment that struggles to connect patterns separated by long gaps in time. The result is a systematic weakness precisely where forecasters need the most help: high-complexity situations and long prediction horizons.

MAST-LLM, which stands for a Multimodal Adaptive Spatio-Temporal framework enhanced by Large Language Models, addresses these weaknesses in three coordinated stages. The first stage is a temporal alignment phase built on a Mixture-of-Experts architecture. A Mixture-of-Experts model is a neural network design in which multiple specialized subnetworks, the experts, each process the input, and a gating mechanism learns to weight their contributions depending on the character of the data at hand. In MAST-LLM, this design lets the framework capture heterogeneous temporal patterns, so that the distinct rhythms of different air traffic variables can each be handled by appropriately specialized experts. Critically, this phase also aligns domain-specific air traffic sequences with the representations that a pre-trained language model has already learned. Rather than forcing raw aviation data into a model built for text, the framework translates traffic dynamics into a form that the language model’s internal representations can meaningfully encode, bridging the gap between two very different data worlds.

The second stage is a full spatio-temporal fine-tuning phase, in which the framework integrates adaptive multimodal spatial representations with multi-scale temporal features. The word multimodal here refers to the combination of different kinds of information about the airspace, such as structural spatial relationships and dynamic traffic measurements, into a shared representation that can adapt as conditions change rather than remaining fixed. The centerpiece of this stage is a mechanism the authors call Bidirectional Spatio-Temporal Attention, or BSTA. Attention mechanisms, first popularized by the transformer architecture underlying modern language models, allow a network to weigh the relevance of every element of its input against every other element. BSTA extends this idea so that spatial and temporal information interact in both directions: spatial structure informs how temporal patterns are interpreted, and temporal evolution informs how spatial relationships are weighted. This joint interaction modeling is what allows the framework to reason about the airspace as a single coupled system rather than as separate spatial and temporal problems stitched together.

The third stage exploits the property that makes large language models attractive for this task in the first place: their capacity for global reasoning and long-range dependency modeling. Language models are trained on sequences in which meaning can depend on context established thousands of tokens earlier, and their architectures are built to preserve and use that distant context. By establishing a unified representation that captures both local dynamics, the minute-to-minute behavior of individual sectors, and global contextual dependencies, the patterns that propagate across an entire region’s airspace over hours, MAST-LLM inherits this long-context strength. The authors argue that this is precisely what earlier frameworks lacked: a way for a prediction about one sector at one moment to draw on evidence from distant sectors and distant times within a single coherent computation.

The empirical results reported in the paper are striking. In extensive experiments, MAST-LLM achieved superior performance at short- and medium-term forecasting horizons and remained competitive at longer horizons, with improvements of up to 9.5 percentage points in accuracy and 12.9 percentage points in macro F1-score over the strongest baselines. The macro F1-score is a particularly meaningful metric here because it averages the F1-score, which balances precision and recall, across all complexity classes, ensuring that improvements on rare but dangerous high-complexity conditions count as much as improvements on routine ones. The authors supplemented the headline numbers with comprehensive ablation studies, which remove individual components to verify that each one contributes, along with factor importance analyses, sensitivity analyses, and visualization analyses that together confirm the effectiveness and robustness of every core element of the design. Statistical significance was assessed with paired two-sided t-tests against the strongest baseline, with exact p-values reported in the paper’s appendix.

The work builds on a substantial lineage of research into both airspace complexity and spatio-temporal machine learning. Measures of air traffic complexity stretch back decades, from early workload prediction studies by Chatterji and Sridhar to probabilistic complexity measures in three-dimensional airspace developed by Prandini and colleagues, and to spatiotemporal graph indicators proposed by Isufaj and collaborators. On the machine learning side, the framework draws on the spatio-temporal graph convolutional networks introduced by Yu, Yin, and Zhu in 2018 and the diffusion convolutional recurrent networks of Li and colleagues from the same year, as well as more recent attention-based architectures such as GMAN and PDFormer. Notably, the same research group had previously developed MAST-GNN, a multimodal adaptive spatio-temporal graph neural network for the same prediction task, and MAST-LLM can be seen as an evolution of that line of work, replacing static graph reasoning with the adaptive, language-model-driven approach.

The study also sits within a rapidly growing movement to apply large language models to time-series and traffic problems. Recent research has shown that pre-trained language models can be reprogrammed for general time-series analysis, as in the One Fits All work of Zhou and colleagues, and for dedicated forecasting frameworks such as Time-LLM by Jin and colleagues and LLM4TS by Chang and colleagues. Parallel efforts have applied these ideas to wind power and wind speed forecasting with BERT4ST, STELLM, and STCA-LLM, to spatio-temporal imputation with GATGPT, and to urban traffic prediction with ST-LLM plus. A 2026 survey by Long and colleagues in IEEE Transactions on Big Data catalogues the accelerating adoption of language models across traffic forecasting applications. MAST-LLM distinguishes itself within this crowded field by combining the Mixture-of-Experts alignment strategy with bidirectional spatio-temporal attention, a pairing the authors present as tailored to the specific structure of airspace dynamics rather than borrowed wholesale from other domains.

The practical implications could be considerable. Accurate complexity forecasts feed directly into demand-capacity balancing, the process by which air navigation service providers decide how much traffic each sector can safely absorb and where flow restrictions should be imposed. Better predictions, especially at medium and long horizons, give managers more lead time to reroute flights, adjust sector configurations, and staff control positions appropriately, potentially reducing both delays and controller overload. The authors have made the airspace complexity dataset used in the study publicly available through a project repository, lowering the barrier for other groups to build on the approach. The research was supported by funders including the National Natural Science Foundation of China and the Ministry of Education of China, and the authors note that generative AI tools were used to assist with language refinement of the manuscript, with all content reviewed and verified by the team.

For the field of air traffic management, the study offers a proof of concept that the reasoning machinery of large language models, originally built for human language, can be redirected toward the physical dynamics of the sky. For the broader machine learning community, it adds to mounting evidence that pre-trained language models serve as powerful general-purpose sequence models whose learned representations transfer far beyond text. Whether such frameworks can be deployed in the safety-critical, certification-heavy environment of real air traffic control remains an open question, and the authors’ results, while strong, come from benchmark evaluation rather than live operations. Still, as global air traffic continues to grow toward and beyond pre-pandemic levels, tools that can anticipate when a sector is about to become unmanageable, minutes or hours in advance, address one of the most consequential prediction problems in transportation, and MAST-LLM demonstrates that the newest generation of AI models may be up to the task.

Subject of Research: Large language model-driven multimodal spatio-temporal prediction of air traffic complexity

Article Title: MAST-LLM: A large language model-driven multimodal adaptive spatio-temporal framework for air traffic complexity prediction

Article References: Cheng, R., Fan, J., Zhang, C., Wu, M., Chen, R., Chai, M., & Wang, A. (2026). MAST-LLM: A large language model-driven multimodal adaptive spatio-temporal framework for air traffic complexity prediction. Applied Intelligence, 56(15), Article 442. https://doi.org/10.1007/s10489-026-07469-7

Image Credits: AI Generated

DOI: 10.1007/s10489-026-07469-7

Keywords: air traffic complexity, large language models, spatio-temporal prediction, mixture of experts, attention mechanism, graph neural networks, air traffic management, time series forecasting, Applied Intelligence, machine learning, airspace sectors, macro F1-score