M.Phil / MS

Permanent URI for this collectionhttps://repository.cuilahore.edu.pk/handle/123456789/36

This collection archives the complete set of theses produced by students of the COMSATS University Islamabad, Lahore Campus.

Browse

Subject: search.filters.subject.fa20

Search Results

Now showing 1 - 8 of 8
  • No Thumbnail Available
    Item
    A Robust Multi-class Skin Cancer Classification using an Attention-based CNN Model
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) Sara Younas; FA20-RCS-015; LHR TP 8352; Dr. Allah Bux Sargano
    Among lesions having high mortality rate, skin lesion is at top of the list. The patient‟s life could be rescued through early diagnosis. The manual identification period could be prolonged, as different categories of skin cancer have high similarities between their structure, color, and size which could lead to misclassification. An automatic and robust system is crucial for the timely and accurate categorization of skin cancer. In the last few decades, deep learning emerges as revolutionizing field, especially in the area of medical imaging. Most state-of-the-art work adopted transfer learning and ensemble learning based techniques for this problem, in which models used for TL and EL are designed for different types of problems and mostly trained on the huge amount of data which sometimes don‟t perform well for distinctive skin cancer problems due to very challenging datasets. This work proposed novel architecture based on a deep learning technique, which is designed especially for skin cancer classification problem. For learning, this method used dermoscopic images, which are excellent quality images captured through a high-level magnifying device, to obtain clear insights of skin containing cancer cells. The proposed architecture‟s main contribution includes, enhancing performance by embedding inception residual (IR) blocks in which, inception block is warily designed with several parallel layers that are merged together. Moreover, residual connections are established in each block to learn from low and high-level features concomitantly. These connections accelerate the learning process along with coping with the vanishing gradient problem. Deployment of the attention unit in the network boosted performance by suppressing the value of noise-containing features and assigning maximum value to important and relevant features during the learning process. Reducing the number of parameters by keenly optimizing the size and no of filters is another milestone attained through suggested model. The presented architecture achieved promising results by attaining 91.63% accuracy, 91.60% x precision, 91.60% sensitivity, 91.52% specificity, and 91.60% f1-score on the ISIC-19 dataset.
  • No Thumbnail Available
    Item
    Deep Learning-Based Prediction of Urban Area Expansion
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) Hamza Ali; FA20-RCS-003; LHR TP 8360; Dr. Zeeshan Gillani
    Multi-temporal data can be useful in many research areas. Multi-temporal data provides us with high-resolution satellite imagery. This high-resolution data can be used to understand change detection in a specific area or the layout of that complete area. To use multi-temporal data, especially custom datasets for deep learning models is real headache. Moreover, Multi temporal data can be very useful with Artificial Intelligence to develop models to tackle different problems like wildfire detection, traffic flow detection, etc. This thesis focuses on two problems 1st one is how can custom multi-temporal dataset which consists of a small sample be used to develop a deep learning model for semantic segmentation purpose of given custom dataset and 2nd problem is related to Urban expansion. The problem of urban expansion is a major issue all over the world, especially in the countries of Africa. Urban expansion has a direct impact on both economic growth and climate change. In this era Machine learning and Computer Vision techniques will provide a vital role to create a model which will help to tackle this problem using multi-temporal data. This thesis methodology is to use custom satellite images data set of a specific area and create a deep learning segmentation model to do segmentation and prediction of different areas in satellite images especially, the urban part. This research focuses on the RGB multi-temporal data set of Dakar, Senegal which is one of the Seaports on the Western Coast of Africa. This thesis methodology is divided into two separate parts. 1st part is about segmentation, an experiment was conducted using with simple Multi U-Net architecture and it achieved more than 85% accuracy on the validation dataset. After these predictions were made using random images from the test dataset. 2nd part focused on the urban expansion problem and trained a Regression model using an Artificial Neural Network which only got a 12% MSE value with 97.25% accuracy and predicts urban expansion on basis of pixel values from remote sensing data using population data from the year 2022 to 2031
  • No Thumbnail Available
    Item
    A Deep Learning Based Prediction of Stock Market Trend using Social Medi
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) Aroma Javed; FA20-RCS-006/; LHR TP 8339; Dr. Hamid Turab Mirza
    Machine learning and deep learning are becoming more and more effective techniques for evaluating financial data, encompassing textual, statistical, and digital information. Future stock prediction is a prominent and challenging deep learning topic in the industry. The difficulty in predicting future stock market stems from too many diverse elements that simultaneously influence the amplitude and frequency of stock market rise and falls. In this research work, the main focus is on the problem of stock market trends predictions using social media as a tool. Digital networks are a fast-growing area of information on the Internet. Perhaps one of the most important features is the instant availability of more knowledge and the users' ability to converse swiftly. Different Deep Learning algorithms (like CNN, RNN, GRU, and Bi Directional RNN) were used to forecast stock market trends based on information from social media, as this data might influence investor behavior. Algorithms were used to investigate the impact of social media accounts on stock market prediction performance. The dataset chosen was an expert and public Twitter post from two prominent technology firms, Alphabet Inc. (Google) and Apple Inc, and news data related to these famous firms. The thesis employed deep learning methods, a pre trained language model for economic sentiment analysis, to extract sentiments from tweets. With the help of this research, it will become easy for an investor to invest his money in companies whose stock market values are high on the basis of sentiment classification and will not lead them to any financial crises. SMP aims to anticipate how the stock value of an economic trade will fluctuate in the foreseeable. If shareholders can precisely estimate stock market progression, investors will indeed be able to turn a profit. Finally, the study predicted the trends by modeling the Data on the proposed GRU model, which outperforms the result of other algorithms. The GRU model has shown significant results with an accuracy of 82.41%.
  • No Thumbnail Available
    Item
    Aspect Based Sentiment Analysis for Urdu Language Using Rule-Based Approac
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) EHTESHAM HASHMI; FA20-RCS-009; Dr. Muhammad Waqas Anwar; LHR TP 8241
    Aspect Based Sentiment Analysis (ABSA) also known as entity level analysis has gained the vital importance to extract the people’s sentiment, emotions or opinions towards some entity. This entity can be any organization, social platforms etc. In spite of the fact that Sentiment Analysis (SA) has been examined broadly within the English language domain with some eminent work in other dialects including Chinese, Arabic etc., many other resource poor languages including Urdu did not get much attention of researchers due to the lack of resources. Urdu is widely spoken language in all over the world and various social platforms are full of Urdu reviews containing people’s sentiments. In this study we have established an aspect level rule-based approach for the Urdu language. To perform our work, we have used the Urdu dialect dataset containing the “COVID-19” tweets. These tweets contain all the information related to Coronavirus and people’s views towards this disease. Sentiment lexicon has been used to extract for the opinion term present in a tweet and after getting the opinion term various aspect have been extracted associated with the opinion. Aspect extraction has been performed using various rule, these rules have been created using linguistic and syntactic context of the phrases present in the tweet and polarities are assigned accordingly. The proposed study focused to achieve four crucial modules: aspect term, aspect term polarity, aspect term category and aspect term category polarity. With the help of evaluation measures including F1-Score, Accuracy, Precision and Recall our work has achieved the promising results.
  • No Thumbnail Available
    Item
    Evaluating Classification of Software Requirements using Machine Learning and Natural Learning Processing Approaches
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) Kinza Tasleem; FA20-RCS-019; LHR TP 7911; Dr. Touseef Tahir
    The software development process consists of a series of phases e.g., requirement engineering, design, coding, and testing, and each phase is critical for fulfilling the needs of a software user. Requirement engineering is vital to understand, analyze and document the needs and expectations of the user. The functional requirements define the roadmap for the software development process. Functional requirements have not gained attention. No state of art discussed formatting and classification of functional requirements subclasses such as ubiquitous. Optional, unwanted behavior, event driven, and state-driven, and there were no larger datasets publicly available furthermore no datasets were formatted in standard syntax. So, the current research focuses to classify functional requirements subcategories e.g, ubiquitous requirements, event-driven, unwanted behavior, optional features, and state-driven requirements. This research aims to format the requirements using the EARS (Easy Approach to Requirement Syntax) boilerplate and perform several DL, ML techniques, and NLP experiments on a larger dataset of more than 9000 requirements which were created through processing 315 software requirement specifications documents of BS (CS) and BS (SE) final year projects (FYP) of CUI, Lahore to classify functional requirements subclasses. Using natural language processing (NLP) and machine learning (ML) techniques, this study intended to create a framework for classifying functional needs and their subclasses. All software requirements were altered through a series of procedures like normalization, and feature extractions techniques like TF-IDF. Several Machine Learning and Deep Learning experiments were conducted e.g., Logistic Regression (LR), Bernoulli Naïve Bayes (BNB), Decision Tree (DT), Multinomial Naïve Bayes (MNB), Random Forest CNN, and Long Short-Term Memory algorithms to classify functional requirements subclasses. CNN model got a higher result about 0.93 and LSTM achieved an accuracy of 0.92
  • No Thumbnail Available
    Item
    Predicting Tourist Destinations based on Interests and Travel Backgrounds using Text Analytics
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) ANAM KHURSHID; FA20-RCS-001; LHR TP 8050; Dr. Hamid Turab Mirza
    Tourist destination prediction has arisen as a new topic within text analytics because of the growing adoption of social media. Tourist data is always beneficial to tourism management since it allows them to give personalized services, products, and destinations to future guests. In predicting tourist destinations, machine learning plays a significant role as it is a way to discover hidden patterns among the dataset; researchers have attempted to get valuable understandings by using data from the real world to train models. Many efforts have been made in this regard; however, most of the models still cannot precisely forecast the tourist’s preferences for destinations. In short, the purpose of this study is to narrow down this gap by introducing a system that uses standard machine learning algorithms to forecast destinations chosen by tourists according to their interests and travel histories. This study extracts a dataset from the CouchSurfing.com website to identify the correlation between tourists' interests and travel backgrounds. The data consists of 9575 records of Pakistan users' profile information. To develop a system with standard accuracy, destinations were divided into 8 regions using Google, and interests were divided into 11 categories using the Yahoo category. Using K-means, clustering has been used several times in this study. This research aims to use classification algorithms to cluster the dataset to identify which interests relate to which countries or regions including KNN, Random Forest, AdaBoost, Gradient Boosting, and LDA. After a comparative results analysis of machine learning algorithms, this research found Gradient Boosting has performed best in classification with the highest accuracy of 99.74% and a kappa Score is 0.99. In the end, the system has successfully predicted tourist destinations related to their interests and past travel experiences. The visitors’ interests according to their destination selections have hardly been studied. This research takes a unique strategy from previous studies in that it focuses on this relationship. For text analysis and future travel destination data, the system has also shown good accuracy and surprising results. As a result, it may help travel companies to build marketing plans for tourists with particular interest categories by offering and advertising places.
  • No Thumbnail Available
    Item
    Contextual Hate Speech over Social Media using Machine Learning and Deep Learning
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2023) Tayyab Rasheed; FA20-RCS-002; Dr. Adnan Ahmad
    With the growing use of social media, information is also increasing dramatically. Because of this growing web material, the disruptions and problems are also increasing. One of the major problems is hate speech, which can cause harmful effects on society, resulting in disputes and chaos. The current automated hate speech detection methods can find the explicit hate very efferently. But now people are using normal words to make hateful content implicit, that is still a challenge for Machine Learning and Deep Learning models. In this research, a dataset is created to detect the context of text in the twister’s hateful tweets targeting others implicitly. Topic modeling is being used for feature extraction and finding out the significance of each word in creating hate. Machine learning and deep learning algorithms along with the proposed feature extraction technique are used to build a context-aware model, that could detect contextual hate speech efficiently.
  • No Thumbnail Available
    Item
    A Robust Multi-Camera Deep Person Re Identification Framework Using Spatiotemporal Context Modelling
    (Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Fatima Zulfiqar; SP20-RCS-005; LHR TP 7601; Dr. Usama Ijaz Bajwa
    In this thesis, research work is based on a robust automatic re-identification of a person from multiple non-overlapping cameras under variable and dynamic environmental conditions for an accurate re-identification and retrieval of targeted person identities. Person Re-Identification (ReID) aims at identifying query person of interest (POI) assigned with a unique identity label across multiple non-overlapping cameras. The query POI can be either an image or a video sequence. Person ReID has gained quite increasing attention among various research and developer communities in recent years. Several research challenges including occlusion, variable viewpoint, misalignment, unrestrained poses, background clutter, etc. are the major challenges in developing robust, lightweight, end-to-end trainable person ReID models. To address these issues, an attention mechanism that comprises local part/region aggregated feature representation learning is presented in this research by incorporating long-range local and global context modeling. The part-aware local attention blocks are aggregated into the widely used modified pre-trained ResNet50 CNN architecture as a backbone employing two attention blocks i.e. Spatio-Temporal Attention Module (STAM) and Channel Attention Module (CAM) thus improving both local and global feature representation learning. The spatial Attention block of STAM can learn contextual dependencies between different human body parts regions like head, upper body, lower body, and shoes from a single frame. On the other hand, the temporal attention modality is capable to learn temporal contextual dependencies of the same person’s body parts across all video frames. Lastly, the channel-based attention modality i.e. CAM can model semantic connections between the channels of feature maps. These STAM and CAM blocks are combined sequentially from a unified attention network named Spatio-Temporal Channel Attention Network (STCANet) that will be able to learn both short-range and long-range global feature maps respectively. Extensive experiments are carried out to study the effectiveness of STCANet on three images and two video-based benchmark datasets i.e. Market- x 1501, DukeMTMC-ReID, MSMT17, DukeMTC-VideoReID, and MARS. K reciprocal re-ranking of gallery set is also applied in which the proposed network showed significant improvement over these datasets in comparison to the state-of-the-art by achieving (mAP/Rank-1) score of (95.5/94.5), (90.7/92.3) and, (74.4/84.5) on Market-1501, DukeMTMC-ReID, and MSMT17 dataset respectively. In addition, the proposed modified STCANet also showed significant performance improvement in comparison to state-of-the-art methods by achieving (mAP/Rank-1) score of (96.6/97.1), (85.3.7/89.1) on DukeMTMC-VideoReID and MARS dataset respectively Lastly, to study the generalizability of STCANet on unseen test instances, cross-validation on external cohorts is also applied that showed the robustness of the proposed model. The proposed STCANet is lightweight, end-to-end trainable, and can be easily deployed to the real world for practical applications
2025 @ COMSATS University Islamabad, Lahore Campus. All rights reserved.