2026 IEEE 2nd International Conference on Quantum Photonics, Artificial Intelligence & Networking (QPAIN)
Article title: “An Explainable Ensemble Framework for Multi-Class Intrusion Detection and Performance Comparison“.
Autors: Md. Shafiqul Islam; Erona Khatun; Md. Momin Hossain; Hasan Khan; Md. Alamgir Hossain; Md. Samiul Islam
Lab: Skill Morph Research Lab., Skill Morph, Dhaka, Bangladesh
Abstract:
Intrusion detection maintains the security of the networks by identifying any attempts to attack or disrupt systems; however, the conventional signature-based techniques often fail to identify new and sophisticated attack patterns. Modern studies indicate that the machine learning methodologies have the potential of identifying threats that were previously unknown. In the paper, we improve an intrusion detection system based on the ensemble learning methods and test the framework with a publicly available dataset. We compare popular models of ensembles, such as Random Forest, Extra Trees, XGBoost, LightGBM, CatBoost, AdaBoost, and Gradient Boosting, in order to determine the most effective one. To advance the transparency of both professionals and amateurs, LIME is used to receive the model outputs, prove that protocol characteristics, transport-layer indicators, and HTTP/DNS header data play a major role in the classification. Across all tests, Random Forest consistently achieves accuracy rates exceeding 99%, coupled with low false positive rates and impressive outcomes in Precision, Recall, F1-score, Balanced Accuracy, and Cohen’s κ. In summary, we present a clear, accurate, and explainable approach to reinforcing computer and network shield to protect against evolving malicious cyber activities.
Article title: “An Interpretable and Effective Machine Learning Method for Stroke Prediction: A Comparison Study Emphasizing the Explainable Extra Trees Classifier’s Superiority“.
Autors: Md. Shafiqul Islam; Sharmin Akter Liza; H.M. Maruf Rahman Shuvo; Md Momin Hossain; Md. Alamgir Hossain; Md. Samiul Islam
Lab: Skill Morph Research Lab., Skill Morph, Dhaka, Bangladesh
Abstract:
To achieve reliable predicted performance, it is necessary to select an appropriate machine learning model, especially in tasks related to classification in which there is an imbalance between classes. To choose the most suitable classifier to solve a binary classification problem, this paper assesses twelve supervised models in a rigorous way. The imbalance in the classes was corrected by applying the Synthetic Minority Oversampling Technique (SMOTE) to the pre-processed dataset (feature scaling and feature encoding). The performance of the model was evaluated using the accuracy, Precision, Recall, F1-Score, and ROC-AUC metrics. XAI approaches were then used to analyze the predictions of the model and the importances of its features to ensure transparency and interpretability. The findings indicate that ensemble-based algorithms are more effective as compared to model families. The Extra Trees and LightGBM classifiers among them gave an accuracy of 94.5 percent. The Extra Trees model was very good with a ROC-AUC of 0.992 and an F1-Score of 94.49 %, as it was very effective in minimizing false negatives. Generally, the Extra Trees classifier was a stable and trusted model that provided balanced performance and high predictive accuracy as can be applied in real-world applications.
Article title: “Two-Stage Noise-Reduced Pneumonia Detection in Chest X-Ray Images using Denoising Autoencoder–CNN Integration“.
Autors: Md Abdul Mutalib; Kopil Das; Mohammad Hossain; Shahriar Hussain; Avijit Deb Nath; Md Mahmud Zaman; Md. Alamgir Hossain
Lab: Skill Morph Research Lab., Skill Morph, Dhaka, Bangladesh
Abstract:
Pneumonia remains a major cause of morbidity and mortality worldwide, with chest X-ray imaging serving as a primary tool for early diagnosis. However, the presence of noise, artifacts, and low-quality acquisitions significantly degrades image interpretability and limits the effectiveness of automated deep learning–based detection systems. Conventional convolutional neural networks (CNNs) are particularly sensitive to such degradations, leading to reduced classification reliability. To address this limitation, we propose a hybrid deep learning framework that integrates a denoising autoencoder (DAE) with a CNN for robust pneumonia detection from chest X-ray images. The DAE is trained to reconstruct clean images from noisy inputs, enabling the suppression of noise while preserving diagnostically relevant features, and is subsequently coupled with a CNN classifier in a unified pipeline. The proposed model is evaluated on the publicly available Kaggle chest X-ray dataset using data augmentation, class-weighted loss, and early stopping to mitigate class imbalance and overfitting. Experimental results demonstrate that the DAE-CNN framework consistently outperforms a CNN-only baseline, achieving an accuracy of 92.9%, an F1-score of 94.6%, and a balanced accuracy of 90.9%, compared to 86.1% accuracy and an F1-score of 89.8% obtained without denoising. These results highlight the effectiveness of incorporating learnable denoising into deep learning pipelines and underscore its potential for improving the robustness and reliability of automated pneumonia detection in clinical imaging scenarios.
Article title: “Secure and Privacy-Preserving Federated Deep Learning Approach for Android Malware Detection“.
Autors: Md. Mostafizur Rahman; Tasnin Khan; Mostak Ahmed; Md. Alamgir Hossain; Md. Samiul Islam; Muhammad Masud Tarek
Lab: Skill Morph Research Lab., Skill Morph, Dhaka, Bangladesh
Abstract:
Android malware has presented serious obstacles to mobile security, which have conventionally been met by centralized machine learning models that invade privacy of users by necessitating the collective compilation of sensitive information. The paper will discuss a Federated Learning (FL) architecture that preserves privacy and uses Deep Neural Networks (DNN) to detect Android malware. With the use of the CIC MalDroid dataset we simulate several clients, each having a mixture of benign applications and different types of malware. Local models are developed on client devices and then combined with the Federated Averaging (FedAvg) algorithm, which realize a promising detection accuracy of 95.86 %. In order to increase the detection capability, local client weights are applied to the global model resulting in local performance improvements and false positives are minimized. The experimental findings prove that the proposed framework is not only highly accurate but also ensures strong generalization among the clients, therefore, the issue of privacy is resolved by storing sensitive data on the local devices.
Article title: “A District Season-Aware Explainable Machine Learning Framework for Diagnosing Climate-Driven Yield Vulnerability in Bangladesh“.
Autors: Ayesha Islam; Md Alamgir Hossain; Md Samiul Islam
Lab: Skill Morph Research Lab., Skill Morph, Dhaka, Bangladesh
Abstract:
Agriculture is a major contributor to Bangladesh’s GDP, with rice being a staple crop crucial for food security. However, climate variability posses major challenges to rice production, which is highly sensitive to temperature and seasonal conditions. This study develops an explainable machine learning (ML) framework for district-level rice yield prediction, integrating long-term climatic variables to uncover underlying climate-yield relationships. Crop yield data of eight major district of Bangladesh were collected from Bangladesh Bureau of Statistics which was combined with NASA POWER reanalysis of climate data of the selected eight district. Eight regression models were evaluated using RMSE, MAE, and R2 metrics, with Gradient Boosting and XGBoost acheving the highest accuracy (RMSE=0.171 and 0.176, respectively). To enhance model interpretability, SHapley Additive exPlanations (SHAP) were applied to the top-performing models. Yield predictions were found to be heavily influenced by temporal trends (year), with minimum temperature and cumulative solar radiation emerging as the most significant climate drivers. Spatial heterogeneity was found in district- and season-wise assessments, with consistently favorable contributions in Chattogram and negative effects in Sylhet. Stronger negative temperature-yield associations were found in Dhaka by comparative analysis, indicating heightened susceptibility to heat stress. The robustness of the identified climate drivers is demonstrated by the similar SHAP patterns across models. The proposed framework provides localized insights into climate and rice yield dynamics, which are essential for policymakers in developing climate-resilient, region-specific agricultural strategies. It can guide resource allocation, risk management, and targeted intervention to improve rice production under changing climatic conditions.
Article title: “Bridging the Precision-Interpretability Gap: A Unified Ensemble Framework for Breast Cancer Detection Enhanced by SHAP and LIME“.
Autors: Hasan Khan; Anika Zaheen; Sharmin Akter Liza; Raju Mia; Obayed ur Rahman Tawhid; Md. Shafiqul Islam
Lab: Skill Morph Research Lab., Skill Morph, Dhaka, Bangladesh
Abstract:
Breast cancer is one of the leading causes of mortality among women in the world where late or false diagnosis compromises death to a large extent. Despite its promising diagnostic power, machine learning has not had high clinical credibility and usage because it cannot be interpreted, which reduces its practical usage. This paper introduces a comprehensible and explicable artificial intelligence-based ensemble model on Breast Cancer Wisconsin Dataset. Different machine learning models have been put to test such as Logistic Regression, Decision Tree, Support Vector Machine, K-Nearest Neighbors, Naive Bayes, Random Forest, Extra Trees, Gradient Boosting, AdaBoost, Bagging and XGBoost and best of them were combined in the form of the soft voting ensemble. The two most successful Logistic Regression and Support Vector machine models are combined with the soft-voting ensemble that is optimized with the use of the GridSearchCV with the ROC-AUC as the parameter of interest. The proposed ensemble had a 98.25 % accuracy, 100% precision, 95.35% recall, 97.62% F1-score, and 99.74% ROC-AUC on the held-out test set. With no incorrectly-determined positives and only two incorrectly-determined negatives (Misclassification Rate: 1.75%). Novel LIME and SHAP methods were used in order to provide explainability and interpretability in diagnosis by global and local explanations. These findings confirm that explainable ensemble models have the potential to provide very precise, interpretable and clinically trustworthy diagnoses of breast cancer, which can ensure safer and more transparent decision-making in the field of women healthcare.
Join skill morph research lab. to learn research and publish research articles.