Dulal Haque

Purpose: Climate unpredictability, soil degradation, and resource scarcity are the main challenges of precision agriculture. Conventional crop recommendation systems lack of real‐time adaptability and transparency. Most machine learning‐based agricultural systems operate as black boxes, hindering farmer trust and adoption. This research addresses these critical gaps by developing an explainable artificial intelligence (XAI)‐integrated, real‐time, portable crop recommendation system that utilize Internet of Things (IoT)‐based soil parameter monitoring with interpretable decision support for precision agriculture. Method: A custom IoT sensor network consists of an RS485 multiparameter soil sensor, an Arduino Uno R3 data logger, and a Raspberry Pi‐5 edge‐computing unit measures seven parameters (N, P, K, temperature, moisture, pH, and electrical conductivity [EC]). Data preprocessing includes K‐nearest neighbor (KNN) imputation, Z ‐score normalization, and synthetic minority over‐sampling technique (SMOTE) augmentation. Twelve conventional machine learning classifiers were initially tested from which four (KNN, random forest [RF], XGBoost, and MLP) were shortlisted for in‐depth benchmarking across 26 crop types. The RF model was selected for edge deployment due to its superior accuracy. A comprehensive explainability framework spanning 12 distinct local and global explanation formats derived from five complementary techniques (permutation importance, SHapley Additive exPlanations [SHAP], local interpretable model‐agnostic explanations [LIME], Morris sensitivity analysis, and explain like i’m 5 [ELI5]) plus counterfactual reasoning provides both local and global interpretability. The model was trained on 31,601 samples and validated on 7901 samples, with inference performed in real time on the edge and exhibited through a deployed web application. Results: RF model had demonstrated the best training accuracy of 99.26% and test accuracy of 98.61%. moisture (0.4207), phosphorus (0.2906), and potassium (0.2848) ranked among the top five in permutation importance. SHAP demonstrated class‐specific visualizations for all 26 crops. LIME estimated the contributions of individual features to the local predictions. The Morris sensitivity was identified as globally sensitive for EC, pH, and temperature. The counterfactual explanations resulted in practical soil amendment recommendations. Conclusion: The first comprehensive XAI‐IoT portable crop recommendation system solves the transparency problem in AI integrated agriculture with 98.61% accuracy. It integrates 12 complementary explainability methods, turning black‐box predictions into acceptable, interpretable decisions, boosting farmers’ confidence and encouraging small farmers to adopt precision agriculture.

Clinical data often exhibit heterogenous feature distributions, nonstationary physiological dynamics, and class imbalance, which collectively increase the complexity of accurate prediction. Despite the popularity of support vector machines (SVMs) in clinical data classification, the traditional kernels have globally fixed hyperparameters and may fail to reproduce the local properties of data. We propose an adaptive density–variance (ADV) kernel for the SVM classifier for addressing these shortcomings. The ADV kernel is based on sample‐conditioned bandwidths that adapt to the local neighborhood density and the feature‐level variance within each sample’s k‐nearest neighbors, directly encoding clinical heterogeneity into the similarity space. It is constructed as a combination of valid kernel Hadamard products (elementwise), an adaptive RBF component, and an interaction term, which is a polynomial. This design explicitly captures localized similarity structures alongside global feature correlations. Furthermore, the resulting kernel matrices behave as an empirically verified positive semidefinite (PSD) via eigenspectrum analysis. This kernel is fully sklearn‐compatible via the BaseEstimator interface (available: https://pypi.org/project/adv-kernel/ ). The proposed approach is tested according to 6 benchmark clinical datasets, including breast cancer, heart disease, PIMA diabetes, fetal health, Parkinson’s disease, and chest X‐ray pneumonia. Clinically relevant metrics are compared between the performance and conventional SVM kernels and standard machine learning models. Statistical significance is determined through Friedman and Wilcoxon signed‐rank tests, and an ablation study is performed to determine the contribution of each kernel component. To overcome the black‐box nature, SHAP‐based explainability and permutation feature importance are also implemented. Nested cross‐validation does offer the reliability of the generalization estimates and unbiased model selection. Experimental findings reveal that the ADV kernel has very competitive performance, especially in F1‐score, balanced accuracy, and agreement measures, which means that it is appropriate in heterogeneous and clinically sensitive healthcare contexts.

The tandem solar cell (TSC) with top and bottom sub‐cells is able to absorb sunlight from visible to near infrared range and is promising for enhancing the photo‐conversion efficiency. We have proposed a TSC with a bandgap of 1.8 and 1.29 eV for carbon nitride (C ₂ N) as a top sub‐cell and tungsten disulfide (WS ₂ ) as a bottom sub‐cell absorber layer, respectively. In this research study, 1D solar capacitance simulation software (SCAPS‐1D) is used to analyze the open circuit voltage (V _oc ), short circuit current (J _sc ), fill factor (FF), and photo‐conversion efficiency (ƞ) of the top and bottom sub‐cell for the design of two‐terminal (2‐T) C ₂ N‐WS ₂ TSC. The current matching condition of J _sc has been determined at the absorber layer thickness of 450 and 790 nm for the top and bottom sub‐cells, respectively. The TSC device parameters such as defect density and doping density have been optimized at 10 ¹⁵ (C ₂ N), 10 ¹⁴ (WS ₂ ) cm ⁻³ , and 10 ¹⁷ cm ⁻³ (C ₂ N and WS ₂ ), respectively for achieving better performance of the proposed structure. The determined optimum values of V _oc , J _sc , ƞ, and FF are 2.38 V, 17.40 mA, 37.60%, and 90.78%, respectively. The present research paves the path for the realization of the high efficiency TSC.

This research explains the advancement and characterization of a highly sensitive hexagonal photonic crystal fiber (H-PCF) optical alcohol sensor in the terahertz range, with an emphasis on improving its capabilities using innovative materials and design. In this instance, the PCF’s core is exposed to the analyte’s flow in order to facilitate detection. This newly developed sensor’s sensing capabilities are carefully investigated using perfectly matched layer (PML) boundary conditions and the fully vectoral finite element method (FV-FEM), facilitated by the COMSOL Multiphysics software. According to the results of the FV-FEM numerical analysis, the proposed optical H-PCF sensor exhibited very high relative sensitivities of 90.65 %, 93.20 %, 93.95 %, and 94.42 % for four alcohols at 1.8 terahertz, namely, ethanol (n=1.354), propanol (n=1.387), butanol (n=1.399), and pentanol (n=1.4072) respectively. Moreover, at 1.8 terahertz (THz), the low confinement losses are 2.11×10-9 dB/m, 9.66×10-11 dB/m, 1.044×10-11 dB/m, 2.66×10-11 dB/m and correspondingly, effective material losses are 0.013497 cm-1, 0.01089 cm-1, 0.010087 cm-1, and 0.009588 cm-1 for similar alcohols. Furthermore, this article provides a brief description of the effective area, effective refractive index, and power fraction. Many contemporary technologies can be used to create this suggested sensor. In addition, let’s anticipate that our sensor will work well for applications in the food and medical sectors to determine the amount of alcohol in different kinds of drinks and beverages. It will also be highly effective in biomedical sectors, public health care, industrial settings, material science, nano-optics, and other terahertz waveguide technology communication domains

Due to the complexity of the fabrication process of tandem solar cells, perovskite materials provide a cost-effective and simpler fabrication solution. In this work, MAPbI₃/MASnI₃ Perovskite-based Tandem solar cell has been explored numerically by using SCAPS-1D. Methylammonium Lead Iodide (CH₃NH₃PbI₃ or MAPbI₃) and Methylammonium Tin Iodide (CH₃NH₃SnI₃ or MASnI₃) have been considered as wide bandgap (WBG) and narrow bandgap (NBG) absorber material for the top cell and bottom cell, respectively. WBG top cell and NBG bottom cell are both investigated through the change in the depth of perovskite materials at the current matching conditions at the illumination of Standard AM 1.5 spectrum. The all-perovskite MAPbI₃/MASnI₃ tandem solar cell with copper (I) oxide (Cu₂O) hole transport layer has achieved excellent performance with a maximum power conversion efficiency (PCE) of 32.74% with V _oc of 1.68 V, J _sc of 24.6 mA cm⁻², and fill factor (FF) of 79.39%, respectively.

This work proposes a hollow-core photonic crystal fiber-based edible oil sensor in the terahertz (THz) range (e.g., 1.0THz ≤ f ≤ 3.0THz) and different sensing characteristics are numerically analyzed. The suggested sensor’s performance was assessed by means of COMSOL Multiphysics, a commercial program that uses the finite element approach. The computational results indicate that the relative sensitivity is 85.591%, 84.648%, 82.625%, 82.683%, and 79.161%, respectively, at f = 2.2 THz, for several types of sunflower oil, mustard oil, coconut oil, olive oil and palm oil; and the corresponding effective areas are 7.22×10-8 um2, 7.09×10-8 um2, 6.83×10-8 um2,7.09×10-8 um2, 6.5231 ×10-8 um2. In addition, the effective material loss for sunflower oil, muster oil, coconut oil, olive oil, and palm oil has been found to be 0.02561 cm-1, 0.027054 cm-1, 0.030322 cm-1, 0.028854 cm-1 ,0.035427cm-1 respectively. Moreover, the proposed sensor also has low confinement loss are 1.55×10-8 dB/m, 1.63×10-8 dB/m, 1.31×10-8 dB/m, 1.99×10-8 dB/m, 4.0345×10-8 dB/m.This proposed sensor can be fabricated using extrusion and 3D-printing technologies, and due to its augmented detecting capabilities, it can be a vital part of oil sensing devices implemented in real life such as industry fields.