PhD
Permanent URI for this collectionhttps://repository.cuilahore.edu.pk/handle/123456789/35
This collection archives the complete set of theses produced by students of the COMSATS University Islamabad, Lahore Campus.
Browse
4 results
Search Results
Item Asset Fractionalization(Library Information Services, CUI Lahore, 2024) MINAL NAZIR; FA20-BCS-071; Dr Wajahat M. QaziWith the cost of living skyrocketing and untrust in financial mechanisms because of one scam after another, for citizens to lay their hands on high valued entity/instrument became a mission impossible. This diminished economic growth, job creation, and less innovation. To solve these problems, the Asset Fractionalization Project redefines accessibility and traceability by fragmentation of physical assets such as real estate, fine art, commodities (e.g., gold), fleets, etc into smaller tradeable units. Thus, investing with small sums and gaining profit from high-value assets will help make it happen. The project's second goal is to implement a framework where investments in fractionalized assets are safe and regulated, where all parties involved — including investors — do not run into fraud or other dangers. Projects like Asset Fractionalization have the potential to revolutionize Pakistan’s financial ecosystem, making it more inclusive and reachable for everyone.Item Automatic Screening of Diabetic Retinopathy at Early Stage using Colored Fundus Images(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Muhammad Nadeem Ashraf; FA14-PCS-003; LHR TP 7864; Dr. Zulfiqar HabibThe research work in this thesis is about an automated diagnosis of diabetic retinopathy (DR) for screening of DR patients at an early stage by classifying both red lesions using color fundus images, through a deep convolutional neural network (DCNN). DR is an asymptotic and vision-threatening retinal complication among working-age adults. Computer-aided diagnosis (CAD) is useful in screening DR patients to assist ophthalmologists to prevent blindness. Researchers have focused on this area by proposing many automated systems to diagnose DR. A comprehensive literature search is conducted during this research. Early-stage diagnosis is challenging due to poor representation of less discriminative and small-sized red lesions of DR. Traditional handcrafted based methods are not usually recommended for clinical trials due to their various limitations. Modern DCNNs are popular to solve various computer vision problems accurately. However, the requirement of huge training data restricts training deep models from scratch to solve the problems in medical imaging. Transfer learning and fine-tuning are helpful alternatives, but overfitting and poor performance further demand the architectural amendments for effective use of the deep CNN models in medical imaging with smaller datasets. Various pre-trained CNN models (AlexNet, VGG16, GoogLeNet, Inception-v3, ResNet50, and DenseNet) are fine-tuned, on the augmented set of 200 × 200 small image patches from the lesion-level annotated images of e-Ophtha_MA dataset, through transfer learning and hyper-parameter tuning to analyze their performance to classify both red lesions of DR. The architecture of a high-performing deep ResNet50 model is selected to fine-tune further. Its baseline architecture is modified by introducing; i) the reinforced skip connections, ii) a Global Max pooling layer, and iii) the Sum-of-Squared Error (SSE) Loss function. The suggested modifications are robust and straightforward. Such alterations may help to enhance various CNN architectures to apply in other imaging domains with small datasets. The proposed framework is evaluated on five publically available datasets: e Ophtha_MA, DiaRetDB1 v2.1, ROC, IDRiD, and Messidor by computing many xi performance metrics. The highest scores (0.9851, 0.991, 0.991, 0.991, 0.991, 0.9939, 0.0029, 0.9879, and 0.9879) of the sensitivity, specificity, AUC, accuracy, precision, F1-score, false-positive rate, Matthews’s correlation coefficient, and kappa coefficient are obtained on unseen test instances from e-Ophtha_MA for DR detection, respectively. The cross-validation results obtained by the proposed method on Messidor and IDRiD datasets at the image level are also better than state-of-the-art techniques. The achieved results are promising and demonstrate the effectiveness of a suggested architecture. For qualitative assessments, the gradient of class activations mapping (Grad-CAM) is computed to visualize the decision for each classified instance. This visual interpretation gives more satisfaction to using the proposed framework for clinical validations. Due to performance, simplicity, and robustness, the suggested model is suitable for diagnosing DR at an early stage for the screening of DR patients. It would help to develop a reliable health care system.Item Spam Review Detection through Behavioral and Linguistic Approaches(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2020) Naveed Hussain; FA15-PCS-002; LHR TP 7730Online reviews regarding different products or services have become the main source to determine public opinions. Consequently, manufacturers and sellers are ex tremely concerned with customer reviews as these have a direct impact on their busi nesses. Unfortunately, to gain profits or fame, spam reviews are written to promote or demote targeted products or services. This practice is known as review spamming. In last few years, the Spam Review Detection (SRD) problem has gained much attention from communities and researchers, but there is still a need to introduce new spam re view detection methods to improve accuracy results utilizing real-world datasets. To overcome these problems, three methods have been proposed. In the first framework, two different spam review detection methods have been pro posed: (i) Spam Review Detection using Behavioral Method (SRD-BM) utilizes thir teen different spammer’s behavioral features to calculate the review spam score which is then used to identify spammers and spam reviews, and (ii) Spam Review Detection using Linguistic Method (SRD-LM) works on the content of the reviews and utilizes transformation, feature selection and classification to identify the spam reviews. Exper imental evaluations are conducted on a real-world Amazon review dataset which ana lyzes 26.7 million reviews and 15.4 million reviewers. The evaluations show that both proposed methods have significantly improved the detection process of spam reviews. Specifically, SRD-BM achieved 93.1% accuracy whereas SRD-LM achieved 88.5% accuracy in spam review detection. Comparatively, SRD-BM achieved better accuracy because it works on utilizing a rich set of spammers’ behavioral features of review da taset which provides in-depth analysis of spammer behavior. Moreover, both proposed methods outperformed existing approaches when compared in terms of accurate iden tification of spam reviews. To the best of the researcher’ knowledge, this is the first study of its kind which uses a large-scale review dataset to analyze different spammers’ behavioral features and linguistic methods utilizing different available classifiers. xi The second method has been developed to analyze the Roman Urdu review dataset based on different classification techniques utilizing linguistic and behavioral features. The performance of each classifier is evaluated in several perspectives: (i) Linguistic features are used to calculate accuracy (F1 Score) of each classifier; (ii) Behavioral features combined with distributional and non-distributional aspects are used to evalu ate accuracy (F1 Score) of each classifier; and (iii) The combination of both linguistic and behavioral features (distributional and non-distributional aspects) are used to eval uate the accuracy of each classifier. The experimental evaluations demonstrated an im proved accuracy (F1 Score: 0.96), which is the result of combinations of linguistic fea tures and behavioral features with the distributional aspect of reviewers. Moreover, be havioral features using distributional characteristic achieved an accuracy (F1 Score) of 0.86 and linguistic features shows an accuracy (F1 Score) of 0.69. The outcome of this research can be used to increase customers’ confidence on online reviews in the South Asian region. It can also help to reduce spam reviews in the South Asian region, partic ularly in Pakistan. The third method proposed Spammer Group Detection (SGD) method, which iden tifies suspicious spammer groups based on the similarity of all reviewer’s activities. Deep learning classifiers are used for training and testing the proposed SGD method. The study also proposed the Diversified Set of Reviews (DSR) method which presents a diversified set of top-k non-spam reviews having positive, negative, and neutral sen timents. Furthermore, it covers all possible features about the product or service. Ex perimental evaluations are conducted on daraz.pk and yelp.com real-world review da tasets. It has been observed by experimental analysis that the proposed SGD method has achieved 89.41% accuracy for the Yelp dataset and 81.31% accuracy for the Daraz dataset in detecting suspicious spammer groups and spam reviews.Item A Digital Twin Assisted Disease Detection for Leafy Green Vegetables Using Federated Learning in Smart Greenhouse(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2025) Iqra Pervez; CIIT/SP23-RCS-016/LHR; Dr. Tariq Umer; LHR TP 9703Agricultural production is seriously vulnerable to the rapid evolution of plant diseases in greenhouse environment. In modern agriculture, ensuring timely and accurate disease detection in leafy green vegetables is crucial for improving crop yield and sustainability. The effects of environmental temperature, humidity on plant’s health which restricts the efficacy of current disease management techniques in greenhouses. This research presents a Digital Twin-assisted model for disease detection in lettuce plants, leveraging Federated-based Learning to enable privacy-preserving and distributed model training within a smart greenhouse environment. The study integrates five pre-trained models AlexNet, MobileNet, ResNet50, EfficientNetB3, and VGG16 to evaluate their effectiveness in classifying plant diseases. This study utilise a lettuce diseases dataset from Kaggle. Among the tested models, VGG16 achieved the highest accuracy of 98%, demonstrating its superior capability for disease classification in resource-constrained settings. Additionally, a correlation analysis between humidity, temperature and disease occurrence was conducted using Unity 3D and Microsoft Azure, providing deeper insights into the environmental conditions influencing plant health. The proposed model is useful for monitoring crops in real time and shows that combining Digital Twin technology with Federated-based Learning can improve decision-making in smart farming systems.