http://hdl.handle.net/123456789/17504 Sat, 17 Oct 2026 01:58:32 GMT 2026-10-17T01:58:32Z http://hdl.handle.net/123456789/21737 SKYLYTICS AN AI POWERED FLIGHT DELAY PREDICTION SYSTEM 03-134222-003, ABDUL MUTLIB ALI Flight delays cost the aviation industry over $28 billion annually, yet most systems remain reactive, and existing research often fails to translate predictive models into real-world operational tools. Skylytics addresses this by providing a full-stack, proactive flight delay prediction system designed for both operational managers and passengers. It uses a hybrid stacking ensemble combining XGBoost and Bidirectional LSTM models, integrated through a Logistic Regression meta-learner, leveraging their complementary strengths. The system is trained on around 14 million flight records (2024–2025), enriched with weather data from Open-Meteo and live air traffic data from OpenSky Network. Multiple models were evaluated under a leakage-free setup, with XGBoost achieving a ROC-AUC of 0.7935, while the BiLSTM contributed significantly to the ensemble. The system is deployed via a FastAPI backend, a Next.js operations dashboard, and a SwiftUI iOS app for passenger tracking. It also includes. SHAP-based interpretability and an AI assistant for natural-language insights. Overall, Skylytics bridges the gap between academic research and production systems, enabling proactive decision-making and improving reliability for both airlines and passengers. DR. NADEEM SARWAR Mon, 01 Jun 2026 00:00:00 GMT http://hdl.handle.net/123456789/21737 2026-06-01T00:00:00Z http://hdl.handle.net/123456789/21740 Bachelor of Science in Information Technology 03-135222-009, Arooj Ishtiaq The goal of this project is to create a complete digital platform called ‘Skillink' that will access a single ecosystem for connecting students, mentors and recruiters. People typically use several individual platforms to find jobs, find mentors, and acquire new skills in the digital world. Skillink overcomes this by incorporating all these features into a single, easy-to-use mobile app. This report dives deep into the development process of a complete mobile app developed with the Flutter platform and connected to a Supabase database.This system is capable of providing authentication, role based access control, real time database. Administrative controls and manipulation of user management. Additionally, the project includes a data analytics aspect which uses Microsoft Power BI to create customer engagement, app trends visualization. Mentoring and income opportunities Dashboard offers useful information for making decisions and growing platforms. The application is developed following a structured approach: requirement Analysis, design of systems, database modeling, implementation and testing. User management, mentorship system, job application tracking and admin dashboard. Among the main modules are to monitor and control platform activities. The system has been tested by various users and implemented successfully. essential features of login, job applications, mentorship requests and the administrative aspects. The findings suggest that the proposed solution offers an improvement in the use of the system and in its efficiency in relation to the current disorganized platforms. Dr. Junaid Nasir Mon, 01 Jun 2026 00:00:00 GMT http://hdl.handle.net/123456789/21740 2026-06-01T00:00:00Z http://hdl.handle.net/123456789/21735 MediCare: AI-Powered Complete Healthcare Management Platform 03-134222-052, Ahsaan Majeed MediCare is an online healthcare management platform that helps patients from finding a doctor, booking a consultation, ordering medicines, and receiving laboratory test reports. It is a multi-panel full-stack platform. The solution offers 5 role-based panels: Patient, Doctor, Pharmacy, Laboratory, and Administrator, each with their personalized sets of features as per user type. The software utilizes React.js with Tailwind CSS on the front end, whereas Node.js with Express.js is used on the backend with MongoDB as the main database. JSON Web Tokens (JWT) are utilized for patient authentication and authorization. The system utilizes Jitsi Meet for video calls between doctors and patients, uses Cloudinary for storage of medical file and prescription, and has example JazzCash integration. An appointment for a face-to-face or video visit can be made. They can order verified pharmacies by merely uploading their prescription. They can order lab tests at home or in the Lab. All appointments, orders and reports can be accessed on one dashboard easily from the patients. Physicians can schedule their week's schedule, have video visits, prescribe medicines digitally and monitor their income. Pharmacies will be able to monitor and know what medicines are available and if they have been ordered. Labatories can book tests, pick up samples, test and upload the pdf of the report that the patient can then download. The platform-wide oversight is provided by an Administrator panel. It gives permission to different service providers such as a doctor, a pharmacist, a laboratory, etc. that will be used by the patients. This verification layer provides trust and authenticity of all service providers on the platform. The platform also has an AI Symptom Analyzer on the homepage that can take a plain-text description of symptoms and provide the top three most relevant medical specialties and top three most matched doctors in each specialty. The Analyzer will use the BART (facebook/bart-large-mnli) zero-shot classification model through the Hugging Face Inference API as the default classification model with a timeout of 5 seconds, and fall back to Anthropic Claude API if the third-party API is not available, so that the feature is always responsive. The platform was evaluated through functional testing across all five panels, demonstrating correct behavior across the full appointment lifecycle, medicine order workflow, and laboratory test workflow. User feedback indicated high satisfaction with the interface design, booking flow, and video consultation experience. Miss. Rabia Maqsood Fri, 01 May 2026 00:00:00 GMT http://hdl.handle.net/123456789/21735 2026-05-01T00:00:00Z http://hdl.handle.net/123456789/21736 Fed-BalX: Federated Adaptive Intrusion Detection Framework for Cross-Domain IoT Environments Using Explainable Attention Fusion and Dynamic Cost Balancing 03-134222-020, Awais Khan The growth of Internet of Things (IoT) networks has made today's environments vulnerable to complex cross-domain attacks that target critical services. This research introduces a federated adaptive machine learning approach to identify multi-vector attacks using the large-scale CICIoT23 dataset, comprising flow duration, header length, TCP/UDP flags, and time-based features. The data was preprocessed to simulate a decentralized network via Dirichlet non-IID partitioning across edge clients, applying local SMOTE to mitigate extreme class imbalance. Visual exploratory data analysis confirmed the relationship between flow time features and malicious packet lengths. A 1D-CNN and BiGRU deep learning model was trained on the federated network using a new Dynamic Class-Balanced Loss (DCBL) function. The model was evaluated using Macro Recall, False Alarm Rate and PR-AUC. The Fed-BalX global model surpassed centralized models with a 11.21 percent improvement in Macro Recall and an 88.89 percent detection metric for high-priority minority attacks, such as DDoS SlowLoris. With a memory footprint of 0.31 MB and a benign error rate of 0.34 percent, the model is extremely practical for real-time anomaly detection at the IoT edge. Using a custom Attention-Guided Explainability (AGX) module and SHAP, we found that Inter-Arrival Times, payload sizes, and connection state flags were key features in making predictions, which are consistent with volumetric attacks. This method shows the feasibility of secure federated learning for advanced IoT threat detection. The results demonstrate the effectiveness of dynamic cost adjustment on highly skewed traffic, and offer a lightweight, scalable framework for smart threat detection. Finally, Fed-BalX is a step towards robust IoT networks that can adapt to and efficiently detect threats early. Miss. Munaza Sher Fri, 01 May 2026 00:00:00 GMT http://hdl.handle.net/123456789/21736 2026-05-01T00:00:00Z