Browse Articles

Principled XAI analysis of the deep learning-based landslide susceptibility prediction model

Abstract Research on applying machine learning (ML) and deep learning (DL) techniques to landslide susceptibility analysis is widespread, with increasingly accurate analyses through novel models. Predicting landslide susceptibility using ML models involves analyzing relationships between conditioning factors and landslide occurrences. Unlike traditional methods, ML models do not explicitly incorporate geotechnical or hydrological theories, raising concerns about result reliability despite high accuracy. This “black-box” limitation has prompted research applying eXplainable Artificial Intelligence (XAI) algorithms to interpret relationships between conditioning factors (digital elevation models (DEM), forest characteristics, soil properties, and geological features) and landslide susceptibility, thereby validating proposed ML models. In this paper, landslide susceptibility prediction models were developed using 20 conditioning factors and multiple architectures, including Support Vector Machine (SVM), Random Forest (RF), Multilayer Perceptron (MLP), and Convolutional Neural Networks (CNNs). Quantitative performances and XAI outcomes were compared. Specifically, the quantitative evaluation showed that the traditional point-based models (RF, SVM, and MLP) achieved Accuracies of 0.6931, 0.6621, and 0.7034, respectively, while the image-based CNNs achieved a higher Accuracy of 0.7586. Furthermore, regarding Recall—a critical metric for disaster management to minimize false negatives—the CNNs (0.8138) significantly outperformed the RF (0.6345), SVM (0.6276), and MLP (0.6828). These results underscore that capturing spatial context through image-wise inputs is far more effective for landslide susceptibility mapping than conventional pixel-level analysis. Because CNNs process input data differently, Gradient-weighted Class Activation Mapping (Grad-CAM) was applied alongside SHapley Additive exPlanations (SHAP) for CNNs, whereas only SHAP was applied to the other models. Results indicated specific patterns associated with certain conditioning factors in landslide susceptibility prediction. CNNs’ Grad-CAM heatmap effectively illustrated these patterns by treating data as images, improving interpretability and reliability of ML outputs.

Explainable ensemble learning using SHAP for ERP anomaly detection

Temperature regulation of a nonlinear CSTR using a global-guided optimization-based PID framework

Abstract Accurate temperature regulation in nonlinear continuous stirred tank reactors (CSTRs) remains a challenging task due to strong nonlinearities and operating-point sensitivity. Although numerous proportional-integral-derivative (PID) tuning approaches have been proposed, most existing studies primarily focus on nominal operating conditions, often resulting in degraded performance under varying process dynamics. To address this limitation, this study proposes an optimization-based PID with filter (PIDf) tuning framework that enhances consistency and reliability across different operating scenarios. A global-guided search mechanism is incorporated into the optimization process to improve convergence stability and solution quality without increasing computational complexity. The proposed framework is evaluated on a nonlinear jacketed CSTR system under setpoint variations and multiple operating conditions. Its performance is benchmarked against recent metaheuristic optimization methods and classical tuning strategies using time-domain specifications and error-based performance indices. The results indicate that the proposed approach achieves faster settling behavior, reduced overshoot, and improved consistency, while maintaining stable performance across repeated runs. Overall, the performance of the proposed approach is quantitatively assessed using standard time-domain and error-based metrics, providing a systematic evaluation of control quality. These findings highlight the applicability of the proposed framework for temperature regulation in nonlinear chemical processes.

Fatigue-associated gut bacteria in Japanese healthy adults characterized by metagenomic analysis

An interpretable radiomics–machine learning model for early risk stratification of invasive fungal infections in community-acquired pneumonia: a dual-center study

Scythicorhinus vekuai gen. nov. et comb. nov. (Mammalia, Rhinocerotidae) from the Pliocene of Georgia and its implications for the early evolution of Coelodonta and Stephanorhinus

Community-aware biased random walks for community detection in attribute networks

Isolation and functional evaluation of Lacticaseibacillus casei HUMB07381 isolated from traditional khiki cheese as a potential probiotic candidate

Uncovering core regulators of multi-abiotic stress adaptation in Arabidopsis thaliana through integrative meta-analysis and machine learning with RT-qPCR validation

Acute pretrauma ethanol exacerbates PTSD-like phenotype in rats and is reversed by early intranasal ketamine

Evaluation of composites interphase mechanical properties considering CZM and FGM behaviors using single-fiber composite tensile tests

Central and peripheral glutamatergic biomarkers delineate heterogeneous clinical phenotypes in olanzapine-treated schizophrenia

Effects of some surface active ionic liquids on the aqueous solubility, thermodynamic properties and fluorescence behavior of drug diclofenac sodium

Abstract Bio-based surface-active ionic liquids (SAILs) hold considerable promise for pharmaceutical applications, particularly in enhancing the aqueous solubility and bioavailability of poorly water-soluble drugs, such as diclofenac sodium (DFS), a nonsteroidal anti-inflammatory agent classified as BCS Class II. Three distinct (2-hydroxyethyl) amine-based surface-active ionic liquids (SAILs) were synthesized. Subsequently, the interactions between DFS and these synthesized SAILs were investigated using fluorescence spectroscopy at a temperature of 298.15 K. Fluorescence spectroscopy revealed a strong interaction between DFS and SAILs. This was evident from the significant quenching of DFS’s intrinsic fluorescence upon SAIL addition. The association constant and binding sites were determined. Among the tested SAILs, the [2-HEA][Ole] exhibited the strongest interaction with DFS. Furthermore, the solubility of DFS in aqueous SAILs solutions studied at temperature range of (298.15 to 313.15) K was found to increase with increasing SAIL concentration. The solubility data were accurately fitted using the e -NRTL and Wilson models. To gain deeper insights, conductor-like screening model (COSMO) calculations were performed on the studied chemicals. The obtained surface cavity volume ( V ) and dielectric solvation energy from the COSMO calculations provided valuable information about the intermolecular interactions. Finally, thermodynamic analysis using Gibbs and van’t Hoff equations indicated that the dissolution of DFS in these systems is an endothermic process.

Multi-objective optimization of corrugated microperforated panels for broadband sound absorption using adaptive sampling method

Exploring tribo-mechanical behavior of B₄C Reinforced LM25 aluminum matrix composites for structural applications

Hypoxia-induced upregulation of lncRNA SLC9A3-AS1 promotes lung adenocarcinoma progression through the miR-506-5p/ASPH/NOTCH1 pathway

Automated measurement of left ventricular ejection time via contactless suprasternal notch laser vibrometry

Spatiotemporal trends in WBGT and affected population in the MENA region (1951–2021)

Abstract Rising temperatures and changes in other meteorological factors have significantly impacted human thermal comfort worldwide. Among global climate change hotspots, the Middle East and North Africa (MENA) have experienced a particularly rapid temperature rise compared to many other regions. Despite its vulnerability, long-term assessments of spatiotemporal heat stress trends and the affected population in MENA remain limited. This study evaluates changes in thermal stress using the wet-bulb globe temperature (WBGT) index derived from ERA5 meteorological data (0.25° resolution) from 1951 to 2020. The findings revealed that the average WBGT in the MENA region increased by 0.75–2.20 °C when comparing 2011–2020 with the 1951–1960 baseline. The most substantial increase occurred in the Arabian Peninsula, specifically Saudi Arabia, where the rise exceeded 0.4 °C per decade in most regions. Consequently, the annual frequency of “Normal conditions” days decreased by approximately 50 days, while high-risk and extreme heat-stress days increased by a corresponding 40 days across most of the region, particularly in the east. Significant trend analysis reveals a 10-day increase per decade across eastern and western MENA under extreme conditions. This rise has led to an additional 1.23 million people experiencing extreme heat stress for at least one day each year. The findings of this study can be useful to policymakers and researchers concerned with extreme heat in the MENA region.

Robust automated detection of small-scale rainfall-induced landslides in Italy Using SegFormer and high-resolution satellite imagery

The prognostic value of different HPV infection statuses in cervical cancer

Abstract To compare oncological outcomes between HPV-positive and HPV-negative cervical cancer patients in a large Chinese multicenter cohort. This retrospective study included cervical squamous cell carcinoma, adenocarcinoma, and adenosquamous carcinoma. Propensity score matching (PSM, 2:1) was used to balance baseline characteristics. Kaplan–Meier method and Cox proportional hazards regression were applied to compare 5-year overall survival (OS) and disease-free survival (DFS). The proportional hazards assumption was tested. Multicenter heterogeneity was adjusted using center stratification. Missing data were managed by complete-case analysis and multiple imputation sensitivity analysis. A total of 11,215 eligible patients were included. Before PSM, the HPV-positive group showed significantly better 5-year OS (92.3% vs. 84.3%, P  < 0.001) and DFS (88.0% vs. 79.6%, P  < 0.001) than the HPV-negative group. Multivariable Cox analysis confirmed HPV negativity as an independent risk factor for OS (HR = 1.545, 95%CI 1.091–2.188, P  = 0.014) and DFS (HR = 1.556, 95%CI 1.212–1.997, P  < 0.001). After 2:1PSM, HPV-positive patients still had superior 5-year OS (88.4% vs. 84.3%, P  = 0.004) and DFS (85.3% vs. 79.6%, P  < 0.001). Among 7,998 patients who underwent radical surgery, similar results were observed before and after PSM. HPV-negative status is associated with inferior 5-year OS and DFS and serves as an independent adverse prognostic factor in cervical cancer, including patients treated with radical surgery.