Drivers’ lack of alertness is one of the main reasons for fatal road traffic accidents (RTA) in Iran. Accident-risk mapping with machine learning algorithms in the geographic information system (GIS) platform is a suitable approach for investigating the occurrence risk of these accidents by analyzing the role of effective factors. This approach helps to identify the high-risk areas even in unnoticed and remote places and prioritizes accident-prone locations. This paper aimed to evaluate tuned machine learning algorithms of bagged decision trees (BDTs), extra trees (ETs), and random forest (RF) in accident-risk mapping caused by drivers’ lack of alertness (due to drowsiness, fatigue, and reduced attention) at a national scale of Iran roads. Accident points and eight effective criteria, namely distance to the city, distance to the gas station, land use/cover, road structure, road type, time of day, traffic direction, and slope, were applied in modeling, using GIS. The time factor was utilized to represent drivers’ varied alertness levels. The accident dataset included 4399 RTA records from March 2017 to March 2019. The performance of all models was cross-validated with five-folds and tree metrics of mean absolute error, mean squared error, and area under the curve of the receiver operating characteristic (ROC-AUC). The results of cross-validation showed that BDT and RF performance with an AUC of 0.846 were slightly more accurate than ET with an AUC of 0.827. The importance of modeling features was assessed by using the Gini index, and the results revealed that the road type, distance to the city, distance to the gas station, slope, and time of day were the most important, while land use/cover, traffic direction, and road structure were the least important. The proposed approach can be improved by applying the traffic volume in modeling and helps decision-makers take necessary actions by identifying important factors on road safety.

The reduction of population concentration in some urban land uses is one way to prevent and reduce the spread of COVID-19 disease. Therefore, the objective of this study is to prepare the risk mapping of COVID-19 in Tehran, Iran, using machine learning algorithms according to socio-economic criteria of land use. Initially, a spatial database was created using 2282 locations of patients with COVID-19 from 2 February 2020 to 21 March 2020 and eight socio-economic land uses affecting the disease—public transport stations, supermarkets, banks, automated teller machines (ATMs), bakeries, pharmacies, fuel stations, and hospitals. The modeling was performed using three machine learning algorithms that included random forest (RF), adaptive neuro-fuzzy inference system (ANFIS), and logistic regression (LR). Feature selection was performed using the OneR method, and the correlation between land uses was obtained using the Pearson coefficient. We deployed 70% and 30% of COVID-19 patient locations for modeling and validation, respectively. The results of the receiver operating characteristic (ROC) curve and the area under the curve (AUC) showed that the RF algorithm, which had a value of 0.803, had the highest modeling accuracy, which was followed by the ANFIS algorithm with a value of 0.758 and the LR algorithm with a value of 0.747. The results showed that the central and the eastern regions of Tehran are more at risk. Public transportation stations and pharmacies were the most correlated with the location of COVID-19 patients in Tehran, according to the results of the OneR technique, RF, and LR algorithms. The results of the Pearson correlation showed that pharmacies and banks are the most incompatible in distribution, and the density of these land uses in Tehran has caused the prevalence of COVID-19.

Sea surface temperature (SST) is crucial in ocean research and marine activities. It makes predicting SST of paramount importance. While SST is highly affected by different oceanic, atmospheric, and climatic parameters, few papers have investigated time-series SST prediction based on multiple features. This paper utilized multi features of air pressure, water temperature, wind direction, and wind speed for time-series hourly SST prediction using deep neural networks of convolutional neural network (CNN), long short-term memory (LSTM), and CNN–LSTM. Models were trained and validated by different epochs, and feature importance was evaluated by the leave-one-feature-out method. Air pressure and water temperature were significantly more important than wind direction and wind speed. Accordingly, feature selection is an essential step for time-series SST prediction. Findings also revealed that all models performed well with low prediction errors, and increasing the epochs did not necessarily improve the modeling. While all models were similarly practical, CNN was considered the most suitable as its training speed was several times faster than the other two models. With all this, the low variance of time-series data helped models make accurate predictions, and the proposed method may have higher errors while working with more variant features.

Addressing water scarcity challenges in arid regions is a pressing concern and demands innovative solutions for accurate groundwater potential mapping (GPM). This study presents a comprehensive evaluation of advanced modeling techniques to enhance the precision of GPM. This study, conducted in the Zayandeh Rood watershed, Iran, employed a spatial database comprising 16 influential factors on groundwater potential and data from 175 wells. This study introduced an innovative approach to GPM by enhancing the Random Forest (RF) algorithm. This enhancement involved integrating three metaheuristic algorithms inspired by human behavior: ICA (Imperialist Competitive Algorithm), TLBO (Teaching-Learning-Based Optimization), and SBO (Student Psychology Based Optimization). The modeling process used 70% training data and 30% evaluation data. Data preprocessing was performed using the multicollinearity test method and frequency ratio (FR) technique to refine the dataset. Subsequently, the GPM was generated using four distinct models, demonstrating the combined power of machine learning and human-inspired metaheuristic algorithms. The performance of the models was systematically assessed through extensive statistical analyses, including root mean squared error (RMSE), mean absolute error (MAE), area under the curve (AUC) for the receiver operating characteristic curve (ROC), Friedman tests, chi-squared tests, and Wilcoxon signed-rank tests. RF-ICA and RF-SPBO emerged as frontrunners, displaying statistically comparable accuracy and significantly outperforming RF-TLBO and the non-optimized RF model. The results of the GPM revealed the exceptional accuracy of RF-ICA, which exhibited a commanding AUC score of 0.865, underscoring its superiority in discriminating between different groundwater potential classes. RF-SPBO also displayed strong performance with an AUC of 0.842, highlighting its effectiveness in inaccurate classification. RF-TLBO and the non-optimized RF model achieved AUC values of 0.813 and 0.810, respectively, indicating comparable performance. The outcomes of this study provide valuable insights for policymakers, offering a robust framework for tackling water scarcity challenges in arid regions through precise and reliable groundwater potential assessments.