Dissertations for Informatics (Knowledge and Data Management)

Permanent URI for this collection

https://bspace.buid.ac.ae/handle/1234/488

Browse

Filter results by year or month

Now showing 1 - 20 of 108

Predicting and Interpreting Student Performance Using Machine Learning in Blended Learning Environments in a Jordanian School Context
(The British University in Dubai (BUiD), 0024-06) SALIM, MAHA JAWDAT; Professor Khaled Shaalan
The rise of the Internet has led to the widespread adoption of digital learning platforms, revolutionising the creation, access, and delivery of digital educational resources. These platforms enhance academic performance by fostering collaborative learning environments and generating extensive data from every user interaction. Machine learning algorithms can process large and complex datasets to identify patterns and trends that may not be immediately apparent. By analysing the data generated from these learning platforms with ML techniques, we can uncover detailed insights into student performance. Accurately predicting student performance can help educators tailor teaching methods and interventions to individual needs. This study focuses on predicting and interpreting student performance in a blended learning environment using ML in a Jordanian school context. The primary aim of this research is to employ machine learning models and SHAP (SHapley Additive exPlanations) to predict and understand student performance. A dataset generated by a digital learning platform used by a private school in Jordan is utilised. Various ML algorithms, such as Support Vector Machines, Logistic Regression, K-Nearest Neighbors, Naïve Bayes, Decision Trees, Random Forest, AdaBoost, Bagging, and Artificial Neural Networks are applied to predict student performance. SHAP values are used to interpret these predictions, offering insights into the factors most impacting student outcomes. Key findings indicate that ensemble methods like Random Forest and Bagging outperform other models in predicting student performance, achieving higher accuracy at 95.90% and 95.48%, respectively, as well as balanced precision and recall, which are crucial for accurately identifying both high- and low-performing students. The findings suggest that using these ensemble methods allows for more reliable predictions and better-informed educational strategies. The analysis reveals that individual features, such as engagement with learning materials and worksheets, significantly influence student performance. By understanding these specific factors and their impacts, educators can tailor interventions more effectively to individual needs, thereby enhancing the educational outcomes and supporting personalised learning. The findings underscore the potential of data-driven strategies to enhance educational outcomes and support personalised learning.
Speech-to-speech example based translation: English to Arabic
(The British Univesity in Dubai (BUiD), 2006-10) Mirza, Ali Arshad
ArgDF: Arguments on the semantic web
(The British University in Dubai (BUiD), 2007-02) Zablith, Fouad
Distributed and Flexible workflow coordination using defeasible logic programming
(The British University in Dubai (BUiD), 2007-12) PV, Sakeer
This is an exploration to foretell a future in which traditional approaches for developing workflow management systems can be supplanted by new techniques and emerging technologies. This thesis recommends sets of methodologies for performing multiagent-based distributed and flexible workflow systems. Its objective is to deal with some of the present issues in the traditional workflow system from the business point of view. This thesis proposes that the traditional design of workflow management systems (clientserver) could be replaced by a defeasible logic programming (DeLP) engine-based multiagent platform that is more flexible and can better replicate workflow’s distributed characteristics in the open environment. This thesis presents sets of technologies for enacting multiagent-based distributed and flexible workflow systems. Its purpose is to tackle some of the existing problems in the traditional workflow system from the business point of view. This thesis proposes that the conventional system architecture of workflow management systems (client-erver) could be replaced by a defeasible logic programming (DeLP) engine–based multiagent platform that is more open and collaborative, and can better reflect workflow’s distributed features in the open environment. This system also eliminates the requirement for centralized workflow coordination and proposes to build a flexible and distributed workflow management system using a multiagent system on a java agent development (JADE) framework, where the workflow semantics and business logic are built using DeLP. The main outcome of this research is to provide approaches for utilizing defeasible logic programming methodologies in application development, especially in business applications, where DeLP can contribute much for the automation of business logic. Moreover, a multiagent system on a JADE framework helps to maximize the use of existing process models and tools for automation of business processes. The model implemented as part of this thesis confirms that a workflow framework using DeLP improves the adaptability and decentralization of workflow management.
Infrastructure for Mass Argumentation Support on the Semantic Web
(The British University in Dubai (BUiD), 2008-03) Banihashemi, Bita
In this thesis,I present an OWL ontology for describing arguments and argument schemes. Following the same key principles of the World Wide Argument Web (WWAW) for building a large-scale Web of structured and inter-connected arguments, this ontology provides an infrastructure for mass argumentation support on the Web.First, I describe the OWL ontology which is based on a new reification of the Argument Interchange Format (AIF) and structures arguments according to Walton’s theory of argumentation schemes. Then, I demonstrate how this ontology enables the use of automated Description Logic reasoning over argument structures. In particular, OWL reasoning enables significantly enhanced querying of arguments through automatic scheme classifications, instance classification, inference of indirect support in chained argument structures and inference of critical questions. Finally, I present the implementation of a Web-based system for authoring and querying argument structures in RDF which utilizes the proposed OWL ontology.
Human Computation for Ontology Refinement
(The British University in Dubai (BUiD), 2008-06) Bussanad Alshamsi, Suaad Salem
Ontology plays an important role in enabling the Semantic Web. Ontologies are used to represent a shared and common understanding of a domain of interest. The main bottleneck is knowledge acquisition in a dynamic environment like the Web. There are several projects that attempts to collect commonsense knowledge, but none of them has explored how to present and use this knowledge in real life. This project explores the possibility of using human computation concepts to devise an online game that enables people to contribute to the process of generating Semantic Web ontologies.The project presents the Ontology Re nement System which generates ontologies from the knowledge acquired by the online game. The system is divided into three main components. The rst component is the Online Game which is used as a tool to collect commonsense knowledge from a large community of people. The second component is the Ontology Builder which represents an algorithm used to illustrate how to process and transfer the knowledge collected into a set of statements. This component is the prime contribution of the project. The third component in the system is the Ontology Representation Generator which takes the set of statements produced by the second component and generates OWL ontology les using Jena API. An experiment has been conducted to evaluate the e ciency of the Ontology Re nement System. Three different ontologies were produced by the system from the knowledge collected during the experiment. A survey about the ontologies produced by the experiment has been sent randomly to Internet users to measure the accuracy of these ontologies. The survey results in a good agreement from the Internet users on the ontologies produced. The ontologies that are produced by the Ontology Re nement System are particularly rel- evant to domains in which ontologies change over time such as electronic commerce. Ontologies can be used to maintain a constantly evolving and improving product catalogue. A good product catalogue can make it easier for people to navigate the available products to find information.
An Online Cognitive, Adaptive, Practice based Approach for ITS: An Approach to IT professional certifications
(The British University in Dubai (BUiD), 2008-10) Duweik, Hikmat Jadalla
With the development of WWW technology, web-based Intelligent Tutoring Systems (ITSs) are becoming an important area of research and development. The key benefit of web-based ITS is that it can be accessed by thousands of learners all over the world. Even though many web-based educational systems have been developed recently, most of them are restricted in functionality, do not provide direction and guidance for individualized learning, and thus fail to take advantage of the distributed nature of the Internet.In this thesis, we present an approach to a practice-based adaptable web-based cognitive intelligent tutoring system that has been applied to build a system that prepares learners for the exam of the International Certificate of ICDL official exam for module 1 (Basic Concepts of Information Technology). ICDL has recently become a de facto standard professional certificate in a wide spectrum of jobs. The system can be used for distance learning/ e-learning or for blended learning approach. It consists of two main modes: the traditional learning mode and the intelligent mode. The traditional learning mode facilitates the learning of concepts by providing full lessons with a fair amount of questions divided between pre lesson quizzes, post lesson quizzes and end of module assessment. The intelligent interactive mode allows the learner to practice concept understanding of the subject matter through means of adaptive self-assessment practices/tests. This is in addition to many other system features including tracking, reporting, assessment/ evaluation tools and some e-communication services and methods.Teachers and students have different interfaces, means of communication, and authorities. The current system consists of nine full lessons with more than 400 questions. An experiment has been conduced to test the system on a random sample of learners to test the effect of the intelligent part over the traditional part on learners learning. The ttest results were impressive and showed that students learning gain on average was 8.1% which is about 3 standard deviations higher when intelligent practiced were involved. The approach is promising and could be extended to cover other subjects and examinations such as SAT, GRE, LSAT, and FE, among others.
Multi-Agent Learning of Strategies in Abstract Argumentation Mechanisms
(The British University in Dubai (BUiD), 2009-01) Nemer, Rama
Argumentation has been studied extensively in the field of Artificial Intelligence, however we know very little about its strategic aspects. This thesis aims to contribute to this general problem by examining the behavior of adaptive self-interested agents, in a multi-agent environment,over repeated encounters using game-theoretic techniques. I extended an existing simulation tool to implement argumentation games and used it to run repeated game experiments using combinations of characteristic argumentation games, adapted from literature, and types of adaptive agents under different conditions.The theme used was that of a court setting whereby there is a judge listening to arguments from different agents. Once all arguments have been presented, the judge must make a ruling: i.e decide which arguments are valid and hence which agents win by presenting them. Agents are assumed to be self-interested and adaptive so they may have conflicting preferences about which arguments they want the judge to accept and they can learn di erent strategies in order to achieve goals that reflect those preferences. The results indicate that the agents use a multitude of different strategies to influence the judge and maximize their payoff , thereby revealing different combinations of arguments with different frequencies, depending on the Nash equilibria of the game, the dominance of the pure strategies and the Pareto e fficiency of the pure strategies in a game. These are dependent on aspects inherent in the argumentation game. While truth revelation was a dominant strategy in some games, interestingly in other cases the agents were able to gain a payoff that is higher than that of all the individual Nash equilibria by playing strategies involving combinations of the Nash equilibria. As for the effect of the learning algorithm on the choice of strategy, the results confirm that WPL is biased toward mixed strategies while GIGA is faster in convergence to pure strategy Nash equilibria. The importance of this kind of work lies in the fact that it combines two aspects of multi-agent systems that have been quite separate to-date: argumentation protocols and multi-agent learning in games.
How Global Coordination Can Be Achieved from Local Coordination in Social Networks
(The British University in Dubai (BUiD), 2009-01) Jahedpari, Fatemeh
Coordination in natural and arti cial intelligent systems has attracted the attention of many scientists recently. However, little research has been conducted to understand the relationship between local coordination and global coordination. In this thesis, we are interested in understanding heuristics that humans use in local coordination problems when they attempt to achieve global coordination. Our goal is to use simulation to systematically analyze the inter-play between frequency bias and degree bias and the group's ability to achieve global coordination. Frequency bias refers to situations where people make choices that result in the fewest local conflicts; and the degree bias refers to situations where people avoid conflict with highly connected neighbors. First,we discuss a basic background about networks and di erent parts of an empirical research in coloring problem. After that, a heuristic that models human behavior which biases towards node's degree is described. Then, we explain an improved version of the model which is capable to capture frequency bias.Finally, we investigate the e ects of frequency bias and degree bias together,using our developed model, in di erent network structures.
Studying the Effects of Listening to Quran on Human Mood
(The British University in Dubai (BUiD), 2009-01) Shaikh, Zeeshan K
Many Muslims believe that listening to their holy book Quran, has an effect on their state of mind or mood. Each Surah of Quran emphasizes on a unique topic, warns of wrong doings and provides the correct way of handling a particular situation. A person waiting for a job interview or promotion who often becomes frustrated can listen or read Surahs that may help him feel better. There has been no work done that attempted to study the e ects of Quran on the mood of people. Furthermore, when people are in a tough situation or in a bad mood,they seldom think about listening or reading Quran. In these situations if there is a device with them that can detect their mood and suggest a particular surah, it can help them feel better. This dissertation presents a proof-of-concept mobile application that can run on any Java enabled phone, which takes user's feedback on Surahs, stores the feedback for further analysis and provides a simple suggestion mechanism.
Self-Organization and Multi-Agent Reinforcement Learning for Taxi Dispatch
(The British University in Dubai (BUiD), 2009-02) Alshamsi Omran, Aamena Ali Ahmed
The taxi dispatch problem involves assigning taxis to callers waiting at different locations. An adjacency-based dispatch system currently in use by a major taxi company divides the city(in which the system operates) into regional dispatch areas. Each area has fixed designated adjacent areas hand-coded by human experts. When a local area does not have vacant cabs,the system chooses an adjacent area to search. However, such fixed, hand-coded adjacency of areas is not always a good indicator because it does not take into consideration frequent changes in tra ffic patterns and road structure. This causes dispatch o fficials to override the system by manually enforcing movement on taxis. In this thesis, I apply two different methods separately to solve the problem: (1) a multiagent self organization technique to dynamically modify the adjacency of dispatch areas (2) a multiagent reinforcement learning method to optimize the dispatch policy for each area. I compare performance of each method with actual data from,and a simulation of, an operational dispatch system. The multiagent self organization technique decreases the total waiting time by up to 25% in comparison with the real system and increases taxi utilization by 20% in comparison with results of the simulation without self-organization. Interestingly, I also discover that human intervention (by either the taxi-dispatch offi cials or the taxi drivers) to manually overcome the limitations of the existing dispatch system can be counterproductive when used with a self-organizing system. Furthermore, the proposed multiagent reinforcement learning method decreases the total waiting time by up to 33.5% in comparison with the real system.
Empirical Studies in Computer-Mediated Interest-Based Negotiations
(The British University in Dubai (BUiD), 2009-05) D'souza, Sohan
Negotiations in which participants exchange o ers based on their chosen positions can be extended to include dialogue about their interests. Revelation of negotiators' interests allows them to make more acceptable o ers and perhaps propose possible alternative approaches toward each other's interests, both of which may result in mutually and individually beneficial outcomes. However, it can also expose their strategies, and possibly their dependencies on other negotiators toward the achievement of their goals. Revealing this information can leave them vulnerable to extortion or retribution, but it can also be used to gain sympathy or build a relationship of trust and reciprocity.This dissertation studies human behaviour and performance upon introducing options for goal inquiry and revelation into mediated-protocol negotiation scenarios. Empirical studies were conducted by having human players negotiate over an alternating o er protocol and an interest-based bargaining protocol, on a platform specially adapted for this purpose. The analysis of data from these experiments revealed interesting patterns in the human use of goal revelation,and its effects on individual and social outcomes and likelihood of agreement. The design of the experiments and the development of the experimentation platform lay the groundwork for the further study of goal revelation in mediated negotiations with humans.
On the Cognitive Plausibility of Abstract Argument Evaluation Criteria: The Case of Argument Reinstatement
(The British University in Dubai (BUiD), 2009-06) Madakkatel, Mohammed Iqbal
Classical reasoning such as reasoning based on propositional logic is monotonic in the sense that adding new information does not remove any previously made conclusion. Common sense suggests that the failure of monotonic reasoning is widespread. Many a time, we jump to conclusions and then we correct our conclusions based on further information as it arrives. Reasoning of this kind is called as nonmonotonic reasoning. Recently, the study of nonmonotonic reasoning has appealed to the powerful notion of argument through the proliferation of so-called argumentation systems. A very influential approach to argumentation systems completely abstracts the origin and the internal structure of the arguments. The focus, instead, is on the relationships between these abstract arguments using defeat relations. The defeat structures can be in different forms such as mutual attack (one argument attacks another argument and the attacked argument attacks the attacker), reinstatement (an argument reinstates another argument by defeating the defeater) and a cycle of attack. The obvious question here is to identify which arguments are rejected, accepted and undecided in such defeat structures. Extension-based abstract argument evaluation criteria (also known as extension-based semantics) can be thought of as criteria for making this decision and have been studied in detail in the literature. These evaluation criteria have been mainly developed for obtaining desirable formal or computational properties, largely based on intuition. The cognitive plausibility of such evaluation criteria has mostly been ignored. However, it is crucial to understand the cognitive plausibility of such evaluation criteria if we are to build software agents capable of interacting persuasively with humans through arguments. This study is an attempt to explore the cognitive plausibility of abstract argument evaluation criteria. Cognitive plausibility of abstract argument evaluation criteria is explored by conducting psychological experiments. Scenarios of standard reinstatement (an argument reinstating another argument by defeating the defeater) and the floating reinstatement (two mutually conflicting arguments reinstating another argument by defeating the defeater) are studied in detail. The empirical results show that the notion of reinstatement in abstract argumentation is cognitively plausible by supporting both grounded and preferred semantics. The results also show that the notion of floating reinstatement is cognitively plausible by not supporting the grounded semantics. Lack of a significant interaction between the pattern and the reasoners' preference (or lack of preference) for one of the two mutually conflicting defenders in the tests indicates the existence of a different cognitively plausible notion that cannot be explained using the abstract argument evaluation criteria. The notion is that there is no clear endorsement for both credulous preferred and sceptical preferred semantics in floating reinstatement. The results also suggest that a floating reinstatement has an e ect that is not signi cantly different from that of the standard form. That is, the mutually conflicting nature of defenders does not play any role that undermines the job of reinstating the main argument. Importantly, only partial recovery is achieved in both scenarios of reinstatement and the idea of such a partial recovery is not dealt with in the abstract argumentation theory.
Prenatal and postnatal information support through the web
(The British University in Dubai (BUiD), 2009-11) Al Janahi, Ayesha
Experimental evaluation of multi-agent reinforcement learning in real-world scale-free networks
(The British University in Dubai (BUiD), 2010-01) Al Hashimi, Rashid
Multi-agent reinforcement learning is a common method for optimizing agents' local decision in a distributed and scalable manner. However, the study and analysis of the state-of-the-art multi-agent reinforcement learning (MARL) algorithms have been limited to small problems involving few number of learning agents.The purpose of this project is to conduct an extensive evaluation and comparison of MARL algorithms when used in networks that exhibit the scale-free property. The Internet and the social network of collaboration in science are only few examples of real-world networks that exhibit this property. Toward this goal, we developed a simulator that facilitates studying combinations of MARL algorithms, strategic games and networks with control propagation via tokens. These tokens are considered an opportunity for agents to play. Tokens also initiate a factor of randomness in the environment given its probability distribution over agents. Preliminary experimental results showed a signi cant reaction to the increase of tokens when agents play battle of the sexes in Neural network; the increase in token transfer probability yields a higher reward and a faster conversion.
Predicting User Behavior over the Web
(The British University in Dubai (BUiD), 2010-01) Al Safadi, Amal Adnan
Web Usage Mining is the application of discovering useful patterns from web data using statistical and data mining techniques. It has recently a wide range of applications in E-commerce web site and E-services such as building interactive marketing strategies, Web recommendation andWeb personalization. Due to its importance, the ECML/PKDD conference announced a competition (challenge) where researchers analyze a web-usage data set and attempt to make predictions about user behavior.The purpose of this thesis is to analyze the first problem of ECML/PKDD 2007 challenge and apply web usage mining techniques in order to predict the user navigation behavior, such as the user visit duration and type of visited pages, based on user real historical data. Toward this goal,I applied web usage mining, data preprocessing, and visualization techniques. I also applied different classification algorithms and studied the effect of attribute selection on each classifier performance. The results I report are comparable to the challenge winner and outperform the runner-up on two out of the three challenge problems.
Security in wire/wireless networks: sniffing attacks prevention/detection techniques in LAN networks & the effect on biometric technology
(The British University in Dubai (BUiD), 2010-02) Al-Hemairy, Moh'd Hussain
During the past era, Information Technology made a revolution in R&D. No double internet becomes an essential backbone of all sciences and research nowadays. Accordingly security threats and data banks attacks turn out to be a phenomenon. Thus, granting protection to such crucial information becomes a high demand. While reviewing the latest studies in this area, there are strong signs that attacking information warehouse is the hot topic nowadays. Moreover, preventing attacks to TCP/IP networks and what are the most efficient techniques to protect it, is the most targeted research area for security experts. For instance, what so called the Man-in-the-Middle attack [MiM] and denial of services [DoS] are just some ways of vulnerable attacks to TCP/IP networks, using some tools available free on the internet. They are sniffing the data traffic or causing service denial. In our research, we evaluated the most famous security solutions and classifying them according to their efficiency against detecting or preventing the types of Address Resolution Protocol [ARP] Spoofing attacks. Based on the surprising experimental results done by a previous study in the security lab which proposed an optimal algorithm to enhance their ability against the two famous network attacks; we implemented the proposed algorithm by this study and stimulate the experiment in order to test the algorithm performance. Moreover we studied the vulnerability of such attacks on two of the most famous security devices used for access-restricted sites; which is known as “Biometric devices”. The biometric technology includes fingerprint readers and Eye readers [iRIS scanners] and we’ll show how they are performing against network attacks.
Colon cancer classification using microarray data
(The British University in Dubai (BUiD), 2010-03) Tariq Khan, Saima
A thesis presented on the classification of cancerous and normal tissue samples using microarray data. In treating cancer time is of the essence and early detection can dramatically increase the chances of survival. Imaging techniques, which are the prevalent method of detection and diagnosis, are only useful once the cancerous growth has become visible.However, if techniques that detect cancerous processes at a genetic level are utilized then the cancerous tissues could be identified, and the disease diagnosed much earlier, thus giving a far better prognosis.Therefore, the aim of this thesis is to evaluate the performance of a variety of different classification methods with a particular dataset containing genetic samples of both normal and cancerous biopsies of the colon tissue.A classifier will be recommended which is able to learn the patterns within the microarray data that best determines the classification of the samples.
Methodology for Weighted Social Networks Navigation
(The British University in Dubai (BUiD), 2010-05) Bin Alnaqeeb, Fatima Mohsen
Our world is becoming an increasingly interconnected world. Connection between different people is being expanded dramatically especially after the vast use of technologies in this area. This expansion necessitates a deep analysis to capture the richness of information that these connections contain. Recently, social networks studies have attracted many researchers from different fields due to their common patterns that exist in wide range of real world networks and the exponential growth of social network sites. One of the important problems in studying social networks is network navigation: how to reach a destination node from a source node using minimum information. In this thesis, our goal is to study the e ect of weights in the network navigation and analyze the inter-play between the homophily, node degree, node strength and node continuous degree. We have identi ed three query routing paradigms based on de ning di erent weights for nodes' edges to guide the navigation process through the network. We then have an extensive experimental study of the performance of incorporating weights into the network for different degrees, homophily parameters and different types of networks.
ALNER: ARABIC LOCATION NAMED ENTITIES
(The British University in Dubai (BUiD), 2010-10) KADDOURA, HAITHAM MOHAMAD
This dissertation describes a rule based approach carried out to determine Location Named Entities in Arabic. ALNER, an Arabic Location Named Entities Recognition system, implements the rule based approach and is introduced in this thesis. This research is the first of its type to specialize in Location NER as a stand-alone system from other named entity types. Such dedication on one named entities helps in investigating the performance of comprehensive NER systems. The Named Entity Recognition (NER) task has great influence on various Natural Language Processing (NLP) applications (e.g. Information Retrieval, Question Answering, etc.). Various research works conducted toward building language independent NER systems that will work on any language but very limited work has been done for NER systems to work with Arabic language. It is known that Arabic language has complex morphology as a language which makes the NER task more difficult. Readers will find an overview about the Arabic language morphology and how it is different from other languages. We also highlighted the key challenges in Arabic language for the NER task. In addition, overall presentation about previous work toward Arabic NER is presented. ALNER system using rule-based approach was evaluated and achieved accuracy of 87.27% and further investigation was conducted to study per module effectiveness and contribution.

Browse

Browsing Dissertations for Informatics (Knowledge and Data Management) by Issue Date

Results Per Page

Sort Options