Hatice Citakoglu

This study examined advanced machine learning models to predict key volumetric parameters of stone mastic asphalt (SMA) mixtures incorporating waste tire rubber. The input parameters were rubber aggregate and rubber modifier. The target outputs were air void, voids filled with asphalt, voids in mineral aggregate, and density. A total of 216 asphalt samples were prepared, representing 72 distinct mixture combinations, as three replicate samples were produced for each condition. Five machine learning techniques—Gaussian Process Regression, Support Vector Machine Regression (SVR), Adaptive Neuro-Fuzzy Inference System, Multivariate Adaptive Regression Splines, and Least Squares Support Vector Machines (LS-SVR) were implemented, and models were evaluated using the 5-fold cross-validation method. Model performance was assessed using mean absolute error, root mean square error, determination coefficient, Wilmott’s refined index, NSE (Nash–Sutcliffe efficiency coefficient), overall index, CA (Combined Accuracy), and RRMSE (relative root mean square error) indices. The LS-SVR model was found to achieve the best predictive performance across all four outputs, with an RRMSE value below 10%, an NSE value greater than 0.9, and a CA value closest to zero. Furthermore, the results show that LS-SVR provides the closest fit to the experimental data and the lowest error levels. In contrast, the standard SVR model produced the weakest predictions, with the widest error distributions and highest deviations. Transforming SVR into LS-SVR significantly improved accuracy. However, the use of only two input variables in this study represents a deliberate simplification that limits the generalizability of the model to broad applications. Consequently, this research is primarily positioned as a proof-of-concept investigation to demonstrate the applicability of ML to a controlled laboratory dataset. Furthermore, the lack of external (independent) validation prevents the developed models from being interpreted as universal predictive tools in practical asphalt mix design. Therefore, the results show that the models successfully capture the relationships between rubber content and volumetric properties within the narrow range of laboratory parameters examined in this study. For broader applications, the inclusion of additional inputs and testing with independent datasets are necessary; these steps will allow for stronger inferences about the actual generalization performance of the models.

Reliable spatial estimation of extreme drought intensity represented by the minimum SPI value is essential for water resources management under increasing climate variability. However, conventional assessments often rely on station-based temporal indices, providing limited representation of spatial heterogeneity and topographic controls. This study developed a machine learning framework to estimate the spatial distribution of extreme drought intensity represented by the minimum SPI value across Türkiye using long-term precipitation records from 102 meteorological stations. For each station and SPI accumulation period, extreme drought intensity was represented by the minimum (most negative) SPI value observed over the 1935–2024 record; this minimum SPI value was used as the machine-learning target, while latitude, longitude, and elevation were used as geographic predictors. Seven machine learning models were evaluated; among them, GBM yielded the highest accuracy (R ² = 0.991, RMSE = 0.007), followed by XGBoost and CatBoost. These results indicate that ensemble tree-based methods reproduce the observed spatial pattern of extreme drought intensity across the station network with high accuracy; establishing predictive skill at unmonitored locations would require evaluation under a station-grouped partition. Furthermore, feature importance and SHAP analyses showed that spatial predictors (longitude, latitude, and elevation) were substantially more influential than the time-scale variable, although their relative importance varied among the evaluated models. This framework offers a robust tool for interpretable and spatially explicit drought monitoring in topographically complex regions.

The aim of this study is to apply time estimation approaches developed based on metrics such as earned schedule (ES), earned duration (ED), and effective earned schedule (ES(e)) within the framework of earned value management (EVM) project control methodology and to model the data obtained from these approaches using machine learning (ML) methods. Accordingly, a comparative analysis of different ML methods in terms of time estimation accuracy has been performed. A dataset is created using 18 different estimated time at completion (EAC(t)) time estimation methods based on data obtained from a residential construction project in Türkiye. Using this dataset, estimation models were developed through ML techniques. The performance of six ML models employing 18-time estimation approaches was evaluated using six performance criteria. The findings indicate that the GPR model achieved superior accuracy compared to the other methods, demonstrating the strong potential of ML-based approaches for time estimation in construction projects.

Reliable multi-station precipitation forecasting is challenging due to nonstationarity, noise, and spatial heterogeneity. This paper introduces a hybrid signal decomposition-machine learning benchmarking framework that integrates four decomposition methods (TQWT, MODWT, EWT, VMD) with three learners (Bagging, LSBoost, KNN), yielding twelve hybrid models. These models were rigorously tested across twelve stations in the Kébir Rhumel Basin using eight statistical metrics and distributional diagnostics to assess accuracy, stability, and generalization. Two dominant families emerged: TQWT-based hybrids achieved localized accuracy at four stations, while MODWT-Bagging led at eight stations and delivered the most consistent cross-station performance. MODWT-Bagging achieved R ² = 0.984–0.993 and NSE = 0.981–0.993, with RMSE ranging from 2.64 to 6.34, demonstrating strong predictive skill under varying hydro-climatic conditions. In noise-rich environments, it substantially reduced errors; for example, at El Milia, RMSE dropped from 12.57 (VMD-LSBoost) to 6.03, a ≈ 52% reduction, and improvements of up to 63% were observed at other stations. Its superiority stems from MODWT’s shift-invariance and noise robustness combined with Bagging’s variance reduction. Taylor diagrams and violin plots confirmed centered, compact error structures, while scatter plots verified accurate phase and magnitude tracking. By clarifying how decomposition structure and learner characteristics interact across heterogeneous regimes, this framework fills a key gap in signal decomposition-machine learning model selection. The findings support adaptive hybrid design for early warning, water resource management, and precipitation-driven forecasting systems. Overall, MODWT-Bagging is established as a robust default for complex precipitation modeling, and the proposed framework provides a scalable foundation for next-generation hybrid predictive tools.

This study examines the historical and anticipated effects of climate change on essential hydroclimatic variables (temperature, precipitation, evapotranspiration, and soil moisture) in the Southeastern Anatolia Project (GAP) region of Türkiye, a semi-arid and agriculturally significant basin experiencing heightened water stress. The analysis employs a collection of CMIP6 Global Climate Models (GCM) and integrates three Shared Socioeconomic Pathways (SSP1-2.6, SSP2-4.5, and SSP5-8.5), utilizing statistical bias correction methods such as Delta Change, Quantile Mapping (QM), and Empirical Quantile Mapping (EQM) to improve the regional accuracy of the projections. The ACCESS-CM2 model, validated with data from Türkiye’s Meteorological General Directorate (MGM), was chosen for comprehensive spatial mapping, utilizing Inverse Distance Weighting (IDW) interpolation across seven temporal intervals encompassing past, present, and future periods. The findings indicate a steady increase in temperature and evapotranspiration, especially under high-emission scenarios, with temperature rises above +4 °C and considerable water losses anticipated by century’s end. Soil moisture exhibits a declining tendency, particularly in the southern and eastern regions, signifying increasing drought susceptibility. Precipitation patterns demonstrate significant spatial variability and rising uncertainty, with relative error (RE%) values increasing under SSP5-8.5. Historical data from 1963 to 2022 corroborate these conclusions, indicating a progressive shift towards a warmer and drier regional climate. These observations highlight the importance of climate adaptation strategies and water management in the GAP region. The research provides decision-makers a high-resolution, bias-corrected hydroclimatic dataset.