Thesis (Bachelor of Science in Computer Science)

Permanent URI for this collectionhttps://hdl.handle.net/10361/28437

Browse

Now showing 1 - 20 of 886

Open Access
Predicting a t20 cricket match result while the match is in progress
(BRAC University, 2015-08) Munir, Fahad; Hasan, Md. Kamrul; Ahmed. Sakib; Quraish, Sultan Md.
Data Mining and Machine learning in Sports Analytics, is a brand new research eld in Computer Science with a lot of challenge. In this research the goal is to design a result prediction system for a T20 cricket match while the match is in progress. Di erent machine learning and statistical approach were taken to nd out the best possible outcome. A very popular data mining algorithm, decision tree were used in this research along with Multiple Linear Regression in order to make a comparison of the results found. These two model are very much popular in predictive modeling. Forecasting a T20 cricket match is a challenge as the momentum of the game can change drastically at any moment. As no such work has done regarding this for- mat of cricket, we have decided to take the challenge as T20 cricket matches are very much popular now a days. We are using decision tree algorithm to design our forecasting system by depending on the previous data of matches played between the teams. This system will help the teams to take major decision when the match is in progress such as when to send which batsman or which bowler to bowl in the middle overs. It significantly expands the exposure of research in sports analytics as it was previously bound between some other selected sports.
Open Access
Spatiotemporal analysis of air pollution using advanced machine learning techniques
(BRAC University, 2026) Rahman, Shafin; Islam, Naeem; Rahman, Md. Shoaibur; Hridoy, Md. Moniruzzaman; Alam, Md. Ahasanul; Department of Computer Science and Engineering
This thesis presents a unified framework for spatiotemporal analysis of air pollu- tion using advanced machine learning to enable short-horizon, citylevel forecasting and operational decision support. A leakage-safe, multi-source dataset is curated for 20 cities across Bangladesh and China, integrating pollutant observations with spatiotemporal covariates (e.g., meteorological and contextual signals) to model ur- ban pollution dynamics under heterogeneous conditions. The forecasting task is formulated as multi-output time-series regression over PM2.5, PM10, NO2, SO2, and CO. To capture short-term fluctuations and longer temporal dependencies while exploiting cross-pollutant structure, a multitask CNN–LSTM architecture is de- veloped with a shared feature backbone and pollutant-specific prediction heads. Performance is benchmarked against classical machine-learning baselines (including Random Forest and XGBoost) under cityaware evaluation to assess both accuracy and robustness. To address regional data imbalance, a cross-country transfer learn- ing strategy is evaluated by leveraging representations learned from data-rich source cities to improve forecasting in data-scarce target cities. Forecast reliability is en- hanced via Monte Carlo Dropout to estimate predictive uncertainty, while SHAP and Integrated Gradients provide complementary explanations of feature influence and temporal attribution. Finally, an early-warning episode detection layer converts forecasts into event-oriented alerts and diagnostics to support practical monitoring workflows. Overall, the proposed pipeline delivers more accurate, uncertainty-aware, and interpretable multi-pollutant forecasts suitable for risk-sensitive air quality man- agement in heterogeneous urban environments.
Open Access
Culturally adaptive neural network for detecting cybersecurity vulnerabilities in Bangladeshi web applications
(BRAC University, 2026) Shaolin, Mohosina; Nawar, Fariha; Siddique, Arik Ahmed; Shams, Shaikh Mohammad Ali; Maliyat, Nafisa; Mostakim, Moin.; Department of Computer Science and Engineering
As cyber threats become more complex and frequent, conventional methods for detecting website vulnerabilities, such as rule-based and heuristic approaches, faces significant difficulties, including limited adaptability, high rates of false positives, and a lack of contextual insight. This study presents a predictive model based on neural networks aimed to actively evaluating website security. By applying essential features like security headers, SSL/TLS settings, and SQL injection vulnerabilities, the model detects complex patterns and irregularities, enabling precise identification of emerging threats and vulnerabilities. This approach uses data-driven feature engineering and training with custom neural architectures, for comparison we used random forest and gradient boosting, For explainability we used SHAP followed by evaluation metrics such as precision, recall, and F1-score. Key results show improved accuracy, reduction of false positives, automated monitoring of configurations, and enhancement of resilience against adversarial attacks. Although neural networks show significant potential for transformation, challenges related to transparency, computational demands, and data imbalance are acknowledged. This highlights the necessity for ongoing learning, scalability, and integration with current frameworks, laying the groundwork for robust and adaptable web security strategies.
Open Access
An end-to-end framework for anomaly detection and categorization
(BRAC University, 2025) Islam, MD. Farhan; Islam, Rehnuma; Reza, Syed Rahin; Tasnim, Saifa; Nipu, Anipa Akter; Rahman, Rafeed; Department of Computer Science and Engineering
In this study, we proposed an end-to-end framework for anomaly detection, classification in Industry 4.0 using deep learning models YOLO V8 and ResNet on the MVTec Anomaly Detection(MVTec AD) dataset. The framework is based on defect detection, anomaly localization. The multitask queues in YOLO V8 guarantee both: fast and precise detection in real time, while ResNet primarily suited for classification, complete with top notch precision and recall metrics. The metrics used for evaluation (including AUC, accuracy, precision, recall, F1 score and AP) confirm the good performance of the models. We also provide decision surface visualizations through Grad-CAM and Integrated Gradients that will help you understand some of the decisions made by the model. The YOLO V8 performed optimal on real-time detection tasks and ResNet performed best on classification accuracy, as highlighted through the results. This framework allows for the automation of anomaly detection and the resolution through investigation, unlocking future opportunities for real time anomaly detection and management.
Open Access
A universal photography suggestion system utilizing composition detection, orientation detection, and subject position detection
(BRAC University, 2025-06) Niloy, Iftikhar Shams; Proma, Syeda Mahjabin; Dofadar, Dibyo Fabian; Ahmed, Md. Sabbir; Department of Computer Science and Engineering
Photography is one of the most popular hobby and images are one of the most important content types on social media, and the impact of a photo often hinges on its composition as much as its subject. In response to this, we proposed a system that classifies the compositional structure, detects orientation and subject of a given photo and suggests improvements based on established photography rules. For the classification of the composition, the photo will be categorized into one of five classes(CC, ROT, LL, FIF, PAT). Then, it will determine the orientation of an image. Lastly, this system uses YOLOv8 object detection model to find the objects of a photograph and through logics and conditions the subject is determined. The proposed system will provide the final suggestion based on the three results of the three proposed models. The main goal of the research is to develop a suggestion system that utilizes the detection models built using Deep Learning(DL) algorithms and find the optimal models that will accurately determine the composition, orientation and subject (if any) of a photograph. We have achieved up to 74.34% accuracy in our composition detection model and a minimum of 0.5870 mean square error (MSE) on our orientation detection model. The subject detection conditions capable of properly detecting the subject of an image most of the cases. Our approach aims to assist users in improving their photography skills and elevating the quality of visual content on any media platforms.
Open Access
Real-time aviation anomaly detection and multi-label classification using deep learning on multivariate sensor data
(BRAC University, 2026) Yeasin, Sakib Rayhan; Rahat, Md. Atik Hasan; Nakib, Shafaat Jamil; Mitra, Debjoty; Alam, Md. Golam Rabiul; Reza, Md. Tanzim; Department of Computer Science and Engineering
General aviation records a fatal accident rate of approximately one per 100,000 flight hours, with loss-of-control and stall events remaining leading preventable causes. Existing flight safety systems rely on fixed expert-defined thresholds and cannot detect complex multi-sensor anomaly patterns or identify specific event types in real time. This thesis proposes a lightweight two-stage deep learning framework for real-time aviation anomaly detection and multi-label event classification on raw flight sensor data from the NGAFID General Aviation Training Set. Stage 1 uses an ensemble of two novel Transformer architectures, the Cross-Sensor Patch Transformer (CSPT) and the Hierarchical Cross-Sensor Transformer (HiCST), each incorporating a cross-sensor multi-head attention module that explicitly models inter-sensor dependencies before temporal processing. Root Mean Square ensemble fusion achieves an anomaly-class F1-score of 0.8815, recall of 0.9076, and AUPRC of 0.9523 on a test set with a 339:1 class imbalance ratio. Stage 2 uses MHANet, which introduces per-sensor independent linear projections to classify each anomalous timestep into any combination of ten simultaneous event types, achieving a macro F1 of 0.9563 and subset accuracy of 0.9627. The complete pipeline runs in 6.08 milliseconds per sensor reading on a standard CPU, confirming real-time feasibility. Both stages outperform all established deep learning and classical machine learning baselines while using significantly fewer parameters, demonstrating that domain-aware architectural specialization consistently outperforms general-purpose approaches for aviation safety monitoring.
Open Access
WasteRefine: boundary-aware semantic segmentation of waste materials using a DINOv2 backbone with multi-scale feature fusion decoder
(BRAC University, 2026-04) Khan, Talha Islam; Das, Trisha; Iqbal, Md. Ahnaf; Tawseef, Farhan; Alam, Md. Golam Rabiul; Datta, Nirjhor; Department of Computer Science and Engineering
The rapid increase in world waste production needs smart, data-driven frameworks for efficient material identification and sustainable resource management. Intelligent recycling systems and waste materials spontaneous segmentation often lack behind due to scarcity of proper annotated datasets, visual ambiguities and severe class imbalancement of rare objects. The research aims to propose WasteRefine, utilizing DINOv2 Vision Transformer backbone with boundary aware semantic segmentation and multi scale feature fusion decoder for waste materials. To capture and accumulate the global context, an advanced dense predictive transformer is used consisting top-down fusion of features, Pyramid Pooling Module, Squeeze and Excitation channel attention and boundary composition component, for the proper identification of cluttered, deformed and visually ambiguous waste objects. The paper also introduces WasteRefine dataset consisting of 2,213 annotated images across four different categories: paper, soft plastic, rigid plastic and metal, marking it as the first waste semantic segmentation dataset from Bangladesh which contains visuals across various regions and annotated precisely. The proposed framework is rigorously evaluated on three different dataset WasteRefine, ZeroWaste-F and SpectralWaste (RGB) and assessed across notable published baselines. The ViT-B achieved 96.64 ± 0.16% mIoU on WasteRefine dataset, 61.94 ± 0.84% mIoU on extremely class imbalanced and deformed ZeroWaste-F dataset and 70.73 ± 0.10% FG mIoU on SpectralWaste beating all the published reports. Competitive results of the ViT-S variant with only 25.16M parameters demonstrated efficient parameter count without severe performance degradation.
Open Access
An efficient technique for real-time transformation of 2D to 3D images with GPU using CUDA programming
(BRAC University, 2026-02) Mahmud, Sadat; Mustafa, Md. Rana; Mitra, Ananda; Shanto, Sajjad Hossain; Chowdhury, Mohammad Nazibul Bashar; Alam, Md. Ashraful; Department of Computer Science and Engineering
This thesis presents a complete 2D to 3D reconstruction system designed to run reliably on a low computational powered PC, where GPU memory, host memory, and disk bandwidth impose strict constraints. The pipeline begins with large-scale synthetic data generation from ShapeNet models, producing aligned RGB and depth observations for supervised learning. A ResUNet18 based monocular depth network is trained in LibTorch using a mask-aware objective to promote numerical stability and reduce invalid-depth regions in the predicted maps. To ensure continuous training without data starvation under limited resources, the system is implemented a producer consumer scheduling system design: a producer renders and stages batches to fast local storage, consumers stream and pre-process shards into the training loop, and a destroyer reclaims storage deterministically once a batch is fully consumed. This design bounds disk usage, prevents host RAM accumulation, and decouples rendering from training so the GPU remains saturated even when CPU-side work fluctuates. After inference, predicted camera-centric depth is lifted into explicit 3D geometry using CUDA-accelerated reconstruction, enabling dense point cloud and grid-mesh generation at image resolution with minimal overhead. The system is evaluated using runtime traces (GPU utilization, GPU/host memory, and CPU load) alongside standard depth estimation metrics aggregated across training batches, demonstrating sustained execution, stable memory behavior, and reconstruction-ready depth quality on resource-constrained hardware.
Open Access
TRACER: task-aware risk-adaptive architecture for continual edge learning
(BRAC University, 2026-02) Islam, Md.Hasibul; Fuad, Mir Muhammad; Tanzin, A.B.M. Fahim Hasan; Rahman, Md. Khalilur; Department of Computer Science and Engineering
Current computer-vision architectures are being deployed as long-lived services, especially on edge and on-device platforms, where input distributions change with changes in environment, users, sensors, and class frequencies. In these cases, to achieve sustainable performance, continual learning is required. Also, we need to keep in mind that the process needs to be feasible under strict constraints like latency and memory. Previous experience demonstrates that device-centric measures of deployment efficiency should be used instead of proxy metrics like FLOPs, and that tail latency (e.g., p95) is a more constrained measure of deployment efficiency than mean latency. At the same time, full neural architecture search (NAS) is generally too costly to integrate into a repeated learning loop, motivating restricted, hardware-aware search strategies. This thesis presents TRACER: Task-aware Risk-adaptive Architecture for Continual Edge leaRning, a deployability-oriented continual learning pipeline that keeps a fixed feature backbone and repeatedly selects and adapts a lightweight MLP classifier head. The system follows a restricted design space with a NAS-inspired controller and a Net2Net-optimized evolutionary population so that it can adapt efficiently. The stability between tasks is ensured through risk-aware exemplar rehearsal (high-risk samples are prioritized) and knowledge distillation. Experiments on Split CIFAR-100 (10 tasks x 10 classes) and CIFAR-10 (5 tasks x 2 classes) report class-incremental (CIL) and task-incremental (Task-IL) performance. Our proof-ofconcept implementation has a final CIL mean accuracy of 0.8145 and average forgetting of 0.0341, and TIL has a final mean accuracy of 0.970 on CIFAR-10. Also, in CIFAR-100, we got 0.6136 final CIL mean accuracy, and TIL has a final mean accuracy of 0.9055 with 0.0451 forgetting. These results, which are derived from a single deterministic run, demonstrate that Lagrangian-relaxation-based constraintaware head selection, combined with risk-sensitive stabilization, provides a practical accuracy-feasibility trade-off for continual learning under explicit latency targets.
Open Access
PAMM: pathway-aware masked representation learning for interpretable multi-cancer prediction
(BRAC University, 2026-01) Chowdhury, Chandrima Roy; Rodoshi, Zarrin Tasnim; Surovi, Sumaiya Hossain; Hasan, Labib; Chakrabarty, Amitabha; Department of Computer Science and Engineering
In this thesis, PAMM, a new paradigm of interpretable multi-cancer prediction based on Pathway-Aware Masked Representation Learning is introduced. To tackle the challenge of the ‘Small n, Large p’ of transcriptomics it is our holding that we apply the rigorous seven-stage pipeline of preprocessing (i.e. Log2 transform, ANOVA filter, Lasso regularization and Recursive Feature Elimination) to reduce the original high-noise 57,750 genes in Breast, Lung, GBM, and HC samples to a high-signal feature set. The basic architecture goes beyond the usual deep learning of black boxes by incorporating biologically relevant priors of KEGG 2021 Human library in a self-supervised masking scheme. In contrast to stochastic masking, the pretraining phase of PAMM uses a Pathway-Aware Masking logic where complete sets of functional genes are zeroed, requiring the model to recreate missing biological units and learn complicated inter-pathway relationships. The latent representations of the model are optimized with Optuna, and the statistical robustness is verified with twenty independent iterations, and the latent representation is further interpreted with Single-sample Gene Set Enrichment Analysis (ssGSEA). The resulting visualizations of mean pathway activity indicate that PAMM is able to capture different, clinically viable biological signatures of each cancer type. PAMM provides a clear and very precise diagnostics platform of precision oncology by filling the gap between high-dimensional self-supervised learning and functional biology. Along with closed-set multi-cancer, PAMM is also explicitly tailored to open-set recognition. Through a combined study of softmax confidence and latent space distances from class centroids, the framework can discard samples that do not adhere to any known cancer manifold. This allows the certainty of identifying unknown or non-cancerous gene expression patterns, which is very essential when it comes to a real-life clinical implementation in which unobservable conditions are the norm. This two-fold feature sets PAMM apart from the traditional classifiers and guarantees the accuracy of the diagnosis and its safety
Open Access
ProtReason: a reasoning-based framework for interpretable protein function prediction
(BRAC University, 2025-06) Ayon, Sartiz Alam; Orin, Alvi Sakib; Biswas, Arpon; Fahad Al Shahid; Shahriyer, Shaikh Faiyaz; Sadeque, Farig Yousuf; Department of Computer Science and Engineering
Understanding how protein sequence determines function remains a central challenge in computational biology. While some protein language models have advanced function prediction but most of them produce outputs without any justification or explainability. Protein function can be justified by connecting biological evidence to functional conclusions. We present ProtReason: A reasoning-augmented framework that generates interpretable protein function predictions with structured reasoning traces. In this study, a curated dataset of 87K proteins is constructed which is enriched with protein domain motifs, localization predictions and structural features transformed into reasoning traces linked to functional labels. ProtReason employs a two-stage architecture that first aligns protein sequence embeddings with textual representations and then generates structured outputs including reasoning traces, functional descriptions, and confidence scores. Compared to a sequence-tofunction baseline without reasoning, ProtReason achieves significantly improved BERT F1 scores, demonstrating the benefit of incorporating reasoning prior to function prediction. A systematic ablation study with 16 model variants shows the best design principles: a single unified reasoning path is better than a multi-step chain of reasoning and generating reasoning before function prediction yields superior performance. ProtReason performs competitively on standard benchmarks while providing biologically interpretable explanations with calibrated confidence estimates.
Open Access
Efficacy of multiple curriculum-based large language models in Bangladesh’s education system
(BRAC University, 2026) Rahman, Aumio; Mazumdar, Tanjila; Suzana, Anika Afsara; Sadeque, Farig Yousuf; Department of Computer Science and Engineering
In Bangladesh, the education system follows multiple curricula, each o”ering di”erent perspectives on culture and society. This research aims to evaluate the e”ectiveness of large language models (LLM) to identify biases in textbooks across di”erent curricula in Bangladesh. As the various educational systems di”er, there is a high possibility that students from di”erent backgrounds may have diverse perspectives on culture and society. Henceforth, this study investigates the potential influences embedded in the curriculum that can a”ect young minds. Our goal is to achieve a comprehensive analysis of these biases, expecting insights that could inform curriculum development and promote balanced educational content in Bangladesh’s diverse educational streams.
Open Access
Automated hazard detection for AR/VR Mars terrain navigation using computer vision
(BRAC University, 2026-01) Rhidy, Tasin Ahsan; Rahman, MD Touhidur; Shuvo, Istiak Zaman; Mahmood, Saiyed Mubasshir; Rafi, Abrar Mojahid; Alam, Md. Ashraful; Alam, Md. Golam Rabiul; Tasnim, Sanjida; Department of Computer Science and Engineering
The exploration of Mars brings about a series of challenges that are occasioned by the risky topographical features, random weather patterns, and the fundamental need to have self-driving equipment. The study builds a combined computer vision and immersive technology system to improve the safety of the human astronauts and robot rovers in their navigation on the surfaces of the Martian environment. Our solution is a multi-modal deep-learning system consisting of object detection, semantic segmentation, and monocular depth estimation to generate complete hazard awareness in simulated Mars environments. We use datasets to train terrain classification models that are able to detect important surface features such as rocks, boulders, and potholes as well as other geological features. The system combines a number of deep-learning networks to detect hazards in real-time and locate bounding-boxes, semantic-segmentation, and pixel-level terrain-classification as well as a depth-estimation architecture to give the system spatial information of the Martian terrain. These models are synergistically used to produce an environmental cognition that drives into an AR/VR interface that provides users with visual cues in safe path planning. The AR/VR element converts raw computer-vision data into usable navigation data, and deciphers warnings of hazards and terrain complexity data to the Martian landscape. The initial studies have shown strong detection of varied terrain conditions, and the multi-modal strategy has a great benefit on improving the safety of navigation in comparison to the single-modality systems. The study has been applied to the development of autonomous planetary exploration technologies and created a scalable model of pre-mission astronaut training and rover operation plan.
Open Access
Temporal state-aware unsupervised anomaly detection for industrial control system
(BRAC University, 2026-01) Anik, Khandoker Wahiduzzaman; Ontu, Md. Rakib Hossain; Sohag, Md. Mehedi Hasan; Badhon, Fardin Jahan; Chakrabarty, Amitabha; Department of Computer Science and Engineering
The increasing interrelationship with Information Technology infrastructure between the Industrial Control Systems (ICS) and critical infrastructure has presented advanced cyber-attacks to critical infrastructure, which not only places data security at risk but also threatens the physical safety, to operational continuity. The thesis will visit the issue of creating an efficient, interpretable and computationally efficient intrusion detection system in an ICS environment through a proposed novel LSTM auto-encoder architecture, specifically trained with edge deployment in mind. The article takes a rigorous approach where physics-conscious feature engineering is embraced, deep-learning architecture creation, and thorough assessment of the WADI (Water Distribution) benchmark data. The proposed system achieves a score of 0.7018 in F1 (Precision=0.7196, Recall=0.7149) by performing the dimensionality reduction of 127 sensors to 30 (which is a reduction of 76 per cent), and by adding the zero-crossing-rate features to the frequency-domain analysis, which is drastically higher than more traditional statistical methods, including Isolation Forest (0.58 F1), or the current state-of-the-art methods, including STADN. To be practical, the system is edge-compatible with an inference latency of 1.84ms, a million parameters, and consumes 5.38W of power when run on simulated NVIDIA Jetson Nano hardware, which is a ten-fold faster inference time than graph-based algorithms. Unsupervised approach, which learns only based on normal operational data, helps to detect novel, zero-day attacks, therefore overcoming the limitation of labelled attack data in operational settings. The study establishes that advanced deep-learning systems can be deployed on the tight computational requirements of industrial edge devices, thus creating a reproducible model of secure and real-time secure critical infrastructure protection.
Open Access
Integrating single sign-on within the WebAuthn framework
(BRAC University, 2026-01) Adnan, Asir; Anika, Nafisha Tabassum; Sobahan, Saima; Istiaque, A.J.M; Ferdous, Md Sadek; Department of Computer Science and Engineering
In the world of digital identity, preserving user privacy while maintaining seamless access across platforms has become a challenge. WebAuthn, developed by World Wide Web Consortium (W3C) is mainly a web-based authentication standard. This system enhances security by enabling passwordless login through hardware-based and biometric authentication mechanisms. Another popular approach, to simplify authentication for users across the internet is Single-Sign-On (SSO) which allows a single credential to access multiple services or applications. This way users can get rid of the liability to manage multiple credentials, rather they can rely on only one credential to authenticate in a trusted manner and use that to authenticate in many other websites. Despite the potential of the SSO system, it has not been integrated with the WebAuthn framework till date. Through our research work, we have introduced a system that ensures passwordless authentication via WebAuthn and supports seamless access to service providers through SSO eliminating the requirements of repeated login. Moreover, this system empowers users with the full control over sharing their personal information by selective disclosure mechanism. Security Assertion Markup language (SAML) is used as the federated identity to exchange the authentication assertion securely between identity providers and service providers to enable seamless SSO. Thereby, introducing a new horizon of research on WebAuthn and SSO.
Open Access
Using machine learning to predict optimal erasure coding policies for object storage system in OpenStack Swift
(BRAC University, 2026-01) Ankon, Amio Malakar; Chaki, Boloy; Sayan, Mashrur Shakhawat; Sarker, Debashish; Mukta, Jannatun Noor; Department of Computer Science and Engineering
Erasure coding helps to reduce storage overhead and improve fault tolerance. But the procedure to select an appropriate erasure coding policy is complex, which often involves tradeoffs among various metrics such as access latency, recovery behavior, storage efficiency, etc. Generally, these are handled using static or heuristic-based configurations, which can not account for variations in workload. In the industry, service providers like Ceph do benchmarking based on throughput/latency without considering workload diversity. OpenStack Swift, one of the most widely used open-source object storage systems, supports erasure coding, but it allocates policy selection in a manual, static way - without it being workload-aware. To address this problem, this thesis showcases a data-driven performance modeling framework for erasure coding in object storage systems using machine learning. A structured dataset- ABDS-30k was constructed in a controlled execution of varying workload conditions. It was created on a Swift All-In-One (SAIO) testbed under different erasure coding policies with the addition of failure injections. The dataset collects several empirically observed performance metrics such as read and write latency, tail latency, success rate, and reconstruction time in case of disk failures. For the job of selecting the optimal erasure coding policy, we adopt two distinct Machine Learning paradigms - a regression-based approach of performance modeling and a classification-based approach for direct data-driven policy recommendation. For regression, CatBoost, XGBoost, and Random Forest Regression are used to predict target metrics in a given workload context and failure scenario, and the optimal policy is chosen based on a weighted score. In the classification-based approach, CatBoostClassifier, XGBoostClassifier, and Logistic Regression are used to label workload–policy pairs using an oracle cost function, and the most optimal erasure coding policy is directly recommended. Experimental results show that the regression models have moderate prediction errors for latency and recovery metrics because of the inherently noisy and heavy-tailed nature of the system, but they remain effective in optimal policy recommendation by exhibiting top-1 accuracy with 48.16% in XGBoost Regression and a top-3 accuracy 100% in Random Forest Regression model. Also, the regret mean (0.010 - 1.81) and regret median (0.00216 - 0.089) values showed a very low margin of error. Classification-based approach shows comparatively weaker metrics, indicating that it is not optimal for data-driven policy recommendation. Overall, this shows the feasibility of using ML-based performance modeling as opposed to static and heuristic-based policy selection, which can be later used as a foundation for future control-plane automations.
Open Access
An optimized predictor for patient health records while ensuring HIPAA compliance
(BRAC University, 2026-01) Das, Stanley Matthew; Alam, Ashiqul; Alvi, Arif Jawad; Mostakim, Moin; Nasim, Hamim Ibne; Tanvir, Sifat; Department of Computer Science and Engineering
The increasing digitization of healthcare has led to predictive analytics becoming an essential tool for early risk detection and personalized patient care. This project introduces an optimized predictor for Patient Healthcare Records. This is a microservices-driven, AI-based system architected to analyze patient data while maintaining HIPAA (Health Insurance Portability and Accountability Act), ensuring scalability through Dockerized deployment. The system functions through three main phases: (1) Data processing through Optical Character Recognition (OCR), which extracts text and refines patient data from medical records; (2) Health Risk Prediction utilizing a Hidden Markov Model (HMM) for sequential health analysis and Neural Networks for predictive modeling; and finally, (3) Secure storage and Recommendations where the predictions are organized in a structured PostgreSQL database and accessed via a web/mobile platform built with HTML and CSS. This design guarantees effective, privacy-conscious, and AI-enabled healthcare analytics, delivering real-time insights for healthcare professionals and providing them with a streamlined, scalable, and secure method for health risk prediction, supporting proactive medical decision-making.
Open Access
Design and analysis of a hybrid spiking–deep neural network for energy-efficient object detection
(BRAC University, 2026-01) Azim, Asrar; Sakib, Md. Arif Uz Zaman; Jamal, Md. Rafid; Yousuf, Nabiha Binte; Jhilik, Shihana Sultana; Rahman, Rafeed; Department of Computer Science and Engineering
In autonomous vehicular systems, object detection performs a fundamental task in which real-time processing and energy e!ciency are crucial for execution on edge devices. Faster R-CNN, which are conventional deep neural network based detectors, provide high accuracy but at the same time sustain substantial power consumption because of continuous and computationally intensive operations. This research proposes the design of a hybrid spiking deep neural network that is energy e!cient in object detection by integrating the architecture of SNN (Spiking Neural Network) that uses LIF (Leaky Integrate and Fire) neurons and temporal spike processing into the feature extraction stage, reducing redundant neural activities simultaneously preserving e”ective object detection functionality by utilizing event-driven spiking computation. The framework is assessed on the KITTI dataset which comprises real world scenarios of autonomous driving. Experimental evaluation shows that it achieves a mAP@0.5 of 0.7395, a mAP@0.7 of 0.5926 and a COCO style mAP@[0.50:0.95] of 0.4771 along with high recall and robust localization. Moreover, the hybrid model enhances detection accuracy over regular SNN models and improves recognition of small and distant objects. These results validate that the presented model has a potential and promising aspect for constructing energy e!- cient object detection systems for autonomous and edge based deployment.
Open Access
Mitigating domain shift in skin cancer recognition with squeeze-excitation attention and dropout-consistent federated averaging
(BRAC University, 2026-01) Uddin, Hasnat Rafi; Tias, Abu Hossain Mohammad Abed Hasan; Islam, Raian; Alam, Md. Ahasanul; Department of Computer Science and Engineering
Skin lesion classifiers can achieve high in-domain accuracy yet fail in clinical deployment, where hospitals di!er in imaging devices, protocols, and class distributions and patient data cannot be centralized. This thesis first reproduces three state of the art baselines (RegNetY-32GF, VGG16, and SkinLesNet) on three widely used datasets: HAM10000, ISIC (9-class), and PAD-UFES-20. We then propose DermaNet, an E”cientNet-B3 based model augmented with squeeze–excitation attention and trained with focal loss, realistic augmentation, and structured dropout. In centralized training, DermaNet achieves 87.83% on HAM10000, while performance on ISIC and PAD-UFES-20 remains substantially lower due to domain shift. To quantify real world generalization, we evaluate cross-dataset transfer and observe catastrophic collapse. For example, DermaNet trained on HAM10000 drops to approximately 40% on ISIC and 20% on PAD, while the ISIC-trained DermaNet falls to approximately 14% on HAM. We address privacy and heterogeneity via federated learning across three clients and benchmark all frameworks under a FedAvg baseline (FedSE), which improves robustness relative to cross domain centralized deployment (84.54% / 69.55% / 65.29% on HAM / ISIC / PAD). However, FedAvg with stochastic dropout su!ers from zero dilution. Structurally dropped weights are averaged as true zeros, weakening attention and slowing convergence. Finally, we propose FedMD, a mask-aware aggregation strategy that averages only active (non-dropped) parameters using deterministic client masks, yielding consistent gains over standard FedAvg, On HAM it achieved 85% accuracy, on ISIC it achieved 71% accuracy and in case of PAD-UFES it achieved 75% accuracy.
Open Access
Multi-task learning framework for drug–target interactions and adverse effects prediction
(BRAC University, 2026-01) Mirza, Md. Sabbir Hossain; Chowdhury, Fahmid Hasan; Rafiq, Zakaria Ibne; Dhali, Radito; Talukder, Md. Rakibul Hasan; Alam, Md. Golam Rabiul; Department of Computer Science and Engineering
Drug-target interactions (DTIs) and adverse drug reactions (ADRs) are biological processes that interact closely but are difficult to model together, preventing the investigation of off-target effects and system perturbations that underlie drug safety. We introduce a single deep-learning system to predict both DTI and ADR using diverse molecular and protein representations of drug SMILES sequences and three-dimensional molecular graphs as well as protein sequences and structural features provided by AlphaFold. Curated DTI and drug-ADR data have a common RxNorm identifier that facilitates the cross-task correspondence between these two data.The proposed context-aware multi-task model uses variational auto-encoder bottleneck to both regularize shared latent space and multihead predictors for binary DTI classification and multi-label ADR with strong label imbalance. In contrast to the previous methods where interaction and safety modeling are decoupled or a single-modality evidence is used, in the model, drug-protein-ADR context is learned jointly. At the drug and protein concentrations, cold-start split analysis show decision-relevant predictions, which are calibrated, and good extrapolation to unfamiliar objects. In addition to predictive performance, the pipeline has a modular and reproducible prediction framework between featurization and inference enabling scalable experimentation. It continues the integration of DTI-ADR modeling as a conceptual method of early safety triage, risk-conscious virtual screening and translational drug discovery. Under stringent protein-level cold-start evaluation, the proposed framework achieves strong and stable performance, attaining a DTI AUROC of 91.60%, AUPRC of 85.46%, and F1-score of 78.39%, alongside ADR prediction with a weighted AUROC of 98.80% and weighted AUPRC of 94.54%, consistently reproduced across multiple random seeds.

Browse

Recent Submissions