Nida Kati

Biomass-derived porous carbons are promising electrode materials for sustainable electrochemical energy storage, yet their capacitance depends on coupled precursor and process variables that are difficult to optimize experimentally. Here, we present a source-aware materials-informatics framework for predicting the specific capacitance of biomass-derived carbon electrodes directly from raw precursor descriptors and processing conditions. A literature-curated dataset comprising elemental composition, proximate analysis, lignocellulosic structure, and activation variables was evaluated with support vector regression, Random Forest, a Deep Neural Network (DNN), Light Gradient Boosting Machine, and Extreme Gradient Boosting (XGBoost), together with an ensemble strategy, across seven feature combinations. XGBoost delivered the strongest test performance, reaching R 2 = 0.948 for the most informative feature block. Beyond model ranking, the study emphasizes validation realism by contrasting conventional random splits with source-aware evaluation, showing that random partitions can substantially overestimate apparent generalization. Across models, current density, activation temperature, holding time, ash content, and volatile matter emerged as the most influential predictors. The resulting workflow provides practical guidance for precursor screening, process-window definition, and data-driven design of biomass-derived supercapacitor electrodes. The guidance applies within the precursor and process regime represented by the training corpus; experimental verification is recommended for excursions beyond it.

Rapid electric-vehicle adoption is reshaping urban energy and mobility systems, especially in multi-unit dwellings (MUDs), where concentrated charging in shared parking areas simultaneously stresses distribution transformers and amplifies the consequences of charger faults, battery thermal events, smoke spread, and emergency-access constraints. The central argument of this paper is that grid stress, resident-facing service quality, lifecycle cost, and fire-risk exposure in enclosed residential parking should be governed jointly rather than as four separate problems. To make that argument concrete, we develop an integrated framework that couples stochastic EV adoption, residential charging-behavior simulation, XGBoost demand forecasting, and linear-programming-based optimization for coordinated control, and we evaluate it through 1000 Monte Carlo trials on representative Turkish MUDs. Unmanaged charging triggers transformer overload at about 30% EV penetration, whereas coordinated control reduces peak demand by 44.7% (405 kW to 224 kW) and raises load factor from 0.40 to 0.68. Strict capacity protection exposes a sharp service–quality trade-off, with only 8.9% of users reaching 80% state of charge (SOC) by departure. Smart charging lowers upfront cost by about 55% ($200 vs. $439 per dwelling unit) and yields roughly $306 net present value per unit over ten years. Building on these results, we propose a five-pillar fire-risk-aware governance architecture—coordinated control, interoperability standards, time-of-use pricing, building–utility coordination, and monitoring—that turns coordinated charging into a preventive governance layer for reducing hazardous congestion in enclosed residential charging environments.

Computer-aided drug discovery increasingly depends on virtual-screening workflows that remain reliable under severe class imbalance, chemical redundancy and early-recognition constraints. In this study, we developed a leakage-aware prioritization workflow for two cancer-relevant targets, pyruvate kinase M2 (PKM2) and mitogen-activated protein kinase 1 (MAPK1/ERK2), using the LIT-PCBA benchmark. The workflow combines canonical-SMILES curation, duplicate and label-conflict auditing, scaffold-aware validation, a non-learning nearest-active Tanimoto baseline, imbalance-aware machine-learning models, repeated-seed robustness analysis, isotonic probability calibration, ensemble-disagreement estimation, absorption, distribution, metabolism, excretion and toxicity (ADMET)-aware triage, molecular docking, and residue-level contact analysis. Benchmark enrichment is interpreted alongside calibration, ADMET filtering, docking and residue-contact evidence, rather than as a standalone discovery claim. PKM2 emerged as the clearer machine-learning case, with scaffold-aware tree models improving early recognition beyond the nearest-active similarity baseline and yielding top-ranked candidates supported by calibrated activity scores, ADMET profiles, docking scores, and residue-contact fingerprints. MAPK1 provided a biologically relevant contrast target, where ligand-neighborhood similarity remained competitive and downstream structural triage became more decisive than ligand-based ranking alone. These results support a conservative drug-discovery workflow in which leakage-aware benchmarking, calibration, uncertainty, and molecular-level triage remain visible throughout candidate prioritization.

This study proposes an intelligent diagnostic system based on machine learning and deep learning for stroke detection. The use of artificial intelligence (AI) in healthcare is increasing alongside big data analytics and digitalization. Stroke, a prevalent neurological disease worldwide, can have its mortality and disability rates significantly reduced through early diagnosis. The study utilizes the “Stroke Prediction Dataset” from Kaggle, encompassing 4909 individuals. This dataset includes 12 input features such as age, gender, hypertension, heart disease, and lifestyle factors, along with one output feature indicating stroke status. Data preprocessing steps involved filling missing values with the mean, converting categorical data to numerical format using One-Hot Encoding, applying Min-Max Scaling, and addressing class imbalance with SMOTE. Fifteen different machine learning and deep learning algorithms (e.g., Random Forest, Voting Classifier, Histogram Gradient Boosting, SVM, MLP) were evaluated, with performance measured using accuracy, precision, recall, F1-score, and ROC-AUC metrics. The Voting Classifier achieved the highest performance with 98.5% accuracy and an AUC of 0.99. Tree-based models like Random Forest and Histogram Gradient Boosting also demonstrated high accuracy. Hyperparameter optimization was performed using GridSearchCV and RandomizedSearchCV, while early stopping, regularization, and dropout techniques were applied to prevent overfitting. The findings highlight the superiority of ensemble learning methods over traditional approaches in stroke diagnosis. The study underscores the importance of integrating AI-based clinical decision support systems into healthcare and suggests that model performance could be further enhanced with larger datasets in the future.

This study presents a process development including detailed characterization of hydrochars and activated carbons produced by hydrothermal carbonization and chemical activation methods using apricot kernel shells (AKS). Initially, the AKS were processed by grinding, followed by subjecting them to hydrothermal carbonization at 240ºC for 24, 36, and 48 hours, resulting in three distinct hydrochars. Subsequently, these hydrochars were mixed with KOH for 3 hours and subjected to a carbonization process at 700°C for 1 hour to obtain activated carbons. Various characterization methods such as Brunauer-Emmett-Teller (BET) surface area analysis, Scanning Electron Microscopy (SEM) and Energy Dispersive Spectroscopy (EDS), X-Ray Diffraction (XRD) analysis, and Fourier-Transform Infrared spectroscopy (FTIR) measurements were employed to determine the properties of the activated carbons. The results obtained indicate that the duration of the hydrothermal reaction increases the carbon content and leads to the formation of porous structures. Particularly, the chemical activation process was found to be effective in pore formation, as evident in SEM images. In conclusion, this study provides a detailed description of the characteristic properties of hydrochars, and activated carbons derived from AKS.

Despite attempts to control the spread of human immunodeficiency virus (HIV) through the use of anti‐HIV medications, the absence of an effective vaccine continues to present a significant obstacle. In addition, the development of drug resistance by HIV underscores the necessity for computational drug discovery methods to identify novel therapies. This investigation specifically focused on employing a long short‐term memory (LSTM) variational autoencoder deep‐learning architecture for computational drug discovery in relation to HIV. Our data set comprised simplified molecular input line entry system (SMILES)–encoded compounds, which were used to train the LSTM autoencoder. Remarkably, our model achieved a training accuracy of 91%, with a data set containing 1377 compounds. Leveraging the generative model derived from the training phase, we generated potential new drugs for combating HIV and assessed their interaction with the virus using a previously developed artificial intelligence model. Lastly, we verified the drug likeliness of our computationally generated compounds in accordance with Lipinski's rule of five. Overall, our study presents a promising approach to computational drug discovery in the ongoing battle against HIV.

The aim of this study was to investigate the performance measurement of supercapacitors using the electrochemical properties of cyclic voltammetry (CV). The use of CV is crucial in evaluating the electrochemical performance of supercapacitors and determining the surface area of the catalyst with regard to the fractal properties of the electrode. The study specifically focused on the CV behavior of a supercapacitor formed by a cobalt-doped ceria/reduced graphene oxide (Co-CeO₂/rGO) fractal nanocomposite, and its assessment was conducted using a machine learning (ML) model with the enhanced XGBoost. The model was trained using an experimental open-source dataset. The results showed that the proposed XGBoost model had a superior ability to predict the CV behavior of the supercapacitor, with nearly perfect results for the MAE, RMSE, and R-squared metrics, which are effective at evaluating the performance of regression models. With the successful design of the proposed intelligent prediction model, the study is expected to provide valuable insights into forming novel nanocomposite forms with high accuracy and minimal need for experiments.