Master's Theses

Permanent URI for this collection


Recent Submissions

Now showing 1 - 20 of 1257
  • ItemOpen Access
    An exploration of school-based sexual abuse prevention programming: No Is a Full Sentence program evaluation
    (Faculty of Graduate Studies and Research, University of Regina, 2023-06) Chetty, Taylor Allison; Milne, Lise; Chalmers, Darlene; Collin-Vézina, Delphine; Daignault, Isabelle
    Child sexual abuse (CSA) is a global issue with potentially long-lasting and severe consequences. In response, school-based prevention programs have been developed and proven to be effective at enhancing child knowledge of CSA concepts. The Saskatoon Sexual Assault and Information Centre’s (SSAIC) new program, “No Is a Full Sentence” (NIAFS), provides grade eight students in both Greater Catholic and Public-School Divisions in Saskatoon, Saskatchewan with foundational and complementary education about CSA concepts, healthy relationships, boundaries, bodily autonomy, understanding sexual violence, and more. This thesis research used a program evaluation framework and program fidelity principles to conduct an evaluation of NIAFS implementation. Semi-structed interviews were conducted with seven participants considered key figures in program development and/or delivery: five SSAIC staff members responsible for program development, implementation, and facilitation; and two teachers who assisted in program facilitation. Thematic analysis led to the construction of four main themes: (1) compromise is required in a number of ways for successful program delivery; (2) facilitators must be attuned to the knowledge, needs, and energy of the classroom organism; (3) facilitators must resonate with the content being shared for its effective delivery; and (4) sexual education isn’t 'just about sex' – it is about planting the seeds required to make change through education. This research looked deeper into the nuances of program delivery in the youth sexual education arena, triangulating findings from SSAIC's post-program satisfaction survey findings, which revealed very positive feedback. Recommendations for future NIAFS delivery, social work practice, and research are provided.
  • ItemOpen Access
    Beyond bubble baths and wine: Broadening perspectives on self-care in female social workers in Saskatchewan
    (Faculty of Graduate Studies and Research, University of Regina, 2023-06) Armstrong, Ivy Allyson; Fletcher, Kara; Gebhard, Amanda; Brand, Denise; Delay, David
    A basic interpretive qualitative research design was used to explore the definition and application of self-care with five female social workers in Saskatchewan. These social workers volunteered to provide information about their experience as a social worker and the potential impact of self-care in their life. Data was analyzed using thematic analysis and included analysis with a feminist lens. There were seven main themes: shifting perspectives about the meaning of self-care, communication, relationships, proactivity, workplaces, holistic self-care, and therapy/counselling. Within these themes were sub-themes used to clarify the self-care practices. Findings are discussed in relation to current research on self-care in social work with recommendations for further research, and implications for practice are included.
  • ItemOpen Access
    Perceptions and experiences of leisure-time physical activity among older adults following a heart attack
    (Faculty of Graduate Studies and Research, University of Regina, 2023-07) Sultana, Sabiha; Genoe, Rebecca; Kelsey, Roz; Kulczycki, Cory; Wickson-Griffiths, Abigail
    Leisure has been found to improve later-life well-being and to help people in coping with life changes (Dupuis & Alzheimer, 2008; Michèle et al., 2019). Leisure activities, including leisuretime physical activity, may significantly affect healthy aging and improve health-related quality of life among older persons. However, there is a lack of literature revealing the determining factors of participation in leisure-time physical activity among older adults following a heart attack. The aim of this research was to explore the perceptions and experiences of leisure-time physical activity among older people who have had a heart attack. To obtain participants’ perspectives, a parallel mixed-methods design was used. Data were collected from 10 participants using a survey (Rapid Assessment of Physical Activity questionnaire) to measure leisure-time physical activity, followed by a face-to-face interview. A qualitative descriptive technique was used to guide the qualitative data collection and analysis. SPSS version 25.0 was used to analyze the demographics and the RAPA questionnaire. Thematic analysis was used to analyze the qualitative data. Four main themes, making lifestyle changes after a heart attack, engagement in leisure-time physical activity, perceptions about leisure-time physical activity after a heart attack, and constraints were generated to describe participants’ perceptions and experiences of leisure-time physical activity. Leisure-time physical activity participation after a heart attack was influenced by several motivators which led to engagement in leisure activities. Participants experienced several constraints in engaging in leisure-time physical activity, however, they described different ways of negotiating those constraints.
  • ItemOpen Access
    Using unmanned aerial vehicles to examine how aboveground forest biomass and bat activity are related to three-dimensional canopy structure
    (Faculty of Graduate Studies and Research, University of Regina, 2023-07) Sprott, Adam Harold; Vanderwel, Mark; Brigham, Mark; McDermid, Gregory J.
    Canopies are emergent properties of a large number of individual trees and provide the underlying structure which affects many other components of forest ecosystems. Unmanned aerial vehicle (UAV) data fills an important gap in aerial imaging by offering an affordable and repeatable source of data to assess variation between forest stands across a given landscape. Individual tree crowns can be segmented from canopy height models to provide detailed information about forest and canopy structure, and photogrammetric point cloud data can be used to identify key habitat variables which affect bat activity. A greater density of biomass may reflect higher productivity and greater energetic and material exchange between the atmosphere and biosphere, whereas increased animal activity indicates a healthier ecosystem with greater resiliency and stability. In this thesis I use data on forest canopy structure 1) to estimate aboveground biomass (AGB) and 2) to model bat activity. I developed allometric scaling models to relate field observations of diameter-at-breast-height (DBH) and tree height to UAV derived tree height and crown area to estimate AGB of individual crown segments. Using these estimates, I constructed two Bayesian regression models to examine variation of AGB of large stands throughout a forest in response to either a set of stand structural predictors or a set of topographic predictors. With these models, I found that at the stand level, AGB was closely related to several structural variables including canopy-area weighted height and tree density and that AGB showed modest relationships with topographic variables such as topographic position, elevation and soil moisture. These results suggest that UAV-derived data calibrated with field observations can be effective for estimating forest AGB and studying its variation among stands, and that stand structural characteristics are more strongly related to AGB that topographic variables. I also captured acoustic recordings of bat passes from some of these forest stands and assigned automatically classified bat calls into three echolocation guilds (short-, mid-, and long-range echolocators; SRE, MRE, and LRE respectively). Using a Bayesian generalized linear model, I was able to model the activity of these guilds in relation to canopy structure. SRE activity was more frequent in taller stands, in stands with less canopy cover, and in stands that were further away from the canopy edge; MRE activity was slightly greater in stands with shorter canopies; and LRE activity was more likely to be observed in stands with more canopy cover and in stands closer to the forest edge. Greater activity of all three bat guilds was observed in plots closer to open water sources. My results illustrate how photogrammetric point clouds can identify fine-scale features useful for AGB modelling and important to bat habitat use, demonstrating how useful UAV captured data can be for forest researchers and managers.
  • ItemOpen Access
    Studying the background response of Minihalo for design optimizations
    (Faculty of Graduate Studies and Research, University of Regina, 2023-08) Sajid, Shayaan; Barbi, Mauricio; Kolev, Nikolay; Ouimet, Philippe; VIrtue, Clarence; Floricel, Remus
    The Neutron Detection Characterization Facility also called Minihalo Neutrino Detector is a planned research and development project that will enhance the detection capabilities of lead based neutrino detectors for supernova physics. It will will be used to construct low background He-3 counters for HALO-1kT supernova neutrino detector and will also provide experimental data on cross-sections of ν-Pb interactions at supernova energy scale. The detector will be placed at the SNS Facility in Oak Ridge National Laboratory on Oak Ridge, Tennesse, USA. The SNS Facility produces three ν species from impinging protons at liquid-mecury target. The νμ, ¯νμ, and νe are produced with a flux of approximately 4.3 ×107 cm−2 s−1 at supernova energy scale. Studying these neutrinos at supernova energy scales at SNS will provide necessary data for HALO-1KT supernova neutrino detector. Detecting neutrinos from core-collapse supernovae through detectors like HALO-1kT accurately can provide invaluable information on the explosion mechanism of massive stars which is not fully understood. This is because neutrinos are emitted from the core of a dying star a couple of hours before the star explodes. Therefore, by detecting these neutrinos, we can not only probe into the heart of exploding stars but also develop full 3D models on their complete explosion mechanism. In order to determine an optimal design for Minihalo, GEANT4 simulations of the proposed design are carried out to study how the detector responds to background at the Facility. Accurate fluxes of cosmic muon and gamma-ray backgrounds at the SNS Facility are simui lated and fired onto the detector. The data from the simulations is analyzed using th ROOT package. This work focuses on the energy deposition of cosmic muons and gamma-rays in the scintillators, the optical photon spectra of scintillators and the background neutrons produced inside the lead from muon interactions. The background neutron spectrum is investigated in detail in order to determine the efficiency of the detector and the necessary change to the proposed design are also investigated to increase the efficiency.
  • ItemOpen Access
    Credit card fraud detection using incremental feature learning
    (Faculty of Graduate Studies and Research, University of Regina, 2023-01) Sadreddin, Armin; Sadaoui, Samira; Khan, Shakil; Bais, Abdul
    Detecting credit card fraud is essential and it is one of the most popular payment methods. Credit card fraud can cause huge losses for cardholders. Therefore, so many studies have focused on proposing different standard machine learning methods and limited use of incremental learning to create a robust detective system. None of these studies can solve all the credit card fraud challenges together. The reason is the complicated real-world scenario and data we have in our hands. Some of these challenges are rapid data arrival rate, concept drift which causes model performance to decline over time and data sensitivity which causes a limited amount of instances in hand for training a model. We have proposed a chunk-based credit card fraud detection model which is based on incremental feature learning and transfer learning. Our proposed approach gives our model the capability to adjust its topology to find the near-optimal solution for the problem at hand. Our approach creates submodels per chunk and for the predictive model creation. We use the most relevant sub-models to the current data distribution we have. By doing so, we do not need to store all the transactions and we can avoid the model infinite growth by setting a limit on the number of used sub-models. There are a limited number of datasets for credit card fraud detection available due to the data sensitivity issue. So, we have evaluated our approach using two of the existing datasets: A mid-scale dataset consisting of two days of European cardholders’ transactions in September 2013 and A large-scale dataset consisting of 6 months of transactions in 2019. We have separated each dataset into a different number of chunks to be able to test and train our approach incrementally. We have compared our approach with a static model based on the initial chunk and re-trained on each chunk. Moreover, we have changed the number of sub-models to evaluate its impact on the performance.
  • ItemOpen Access
    Precision-based boosting for regression
    (Faculty of Graduate Studies and Research, University of Regina, 2023-06) Razavi, Mahsasadat; Zilles, Sandra; Mouhoub, Malek; Weger, Harold
    Regression is a type of predictive modeling problem that involves estimating a continuous numerical value based on input variables. The goal of this research is to investigate whether incorporating the precision of regression models on specific target values can improve the performance of ensemble-based regression models. We begin by reviewing two existing ensemble methods for classification, namely AdaBoost and PrAdaBoost, which will form the basis of our proposed ensemble method for regression. We also provide a formal analysis of the training error upper bounds for PrAdaBoost and AdaBoost. The mathematical proof shows that PrAdaBoost’s upper bound is always less than or equal to AdaBoost’s. This result is important because it implies that PrAdaBoost’s training error upper bound decreases exponentially as the number of iterations increases, assuming that each individual predictor in the ensemble is better than random guessing. We modify the PrAdaBoost algorithm and implement it in the context of regression, thus introducing a new regression algorithm called PrSAMME-R. To evaluate the performance of PrSAMME-R, several experiments are conducted on various regression datasets, and the results are compared to those obtained from other ensemble-based regression models. The results show that incorporating the precision of regression models on specific target values into their weights can improve the performance of ensemble-based regression models significantly. PrSAMME-R outperforms other ensemble-based rei gression models such as Random Forest, Gradient-based Boosting, AdaBoost.R2 and AdaBoost.RT, in terms of mean absolute error.
  • ItemOpen Access
    Sword fighting in virtual reality: Where are we and how do we make it real
    (Faculty of Graduate Studies and Research, University of Regina, 2023-03) Pitura, Philip Remo Stanley; Gerhard, David; Hamilton, Howard; Dorsch, Kim
    Virtual Reality has historically been a research space concerned with recreating the natural world around us. It is currently best known for its role in gaming and escapism. A number of different mediums have used virtual reality for its strengths in training. It is particularly useful for its ability to recreate real world locations and situations while still having full control over the environment. One area where it has failed in this realism is in its portrayal of fencing. Fencing, also known as sword fighting, is a common interaction in virtual reality gaming. Virtual reality fencing is plagued by a lack of features that are necessary to achieve realistic fencing. In this thesis I present eleven features gathered through observations of seven games. Weapon weight, parries, edge detection, edge alignment, point detection, weapon flex, blade tracking, response to physical locomotion, quality of expected movements, and weapon interaction with the environment are identified as necessary features in a realistic fencing experience. These features are explored with respect to their appearance in games as well as with regard to the current issues surrounding their implementation. It is found that all eleven features are necessary for the creation of a realistic fencing experience and they will require a physics based approach to their implementation and interactions.
  • ItemOpen Access
    Two-parameter super-product systems of compact Hausdorff spaces
    (Faculty of Graduate Studies and Research, University of Regina, 2023-06) Melanson, Patrick; Floricel, Remus; Argerami, Martín; Zilles, Sandra
    The theory of C∗−algebras [1, 7,17] is incredibly rich and provides a great starting point for exploring various types of operator algebras within the realm of Functional Analysis. Previous papers [8, 12] have used these algebras to analyze what are called C∗-product systems and C∗-subproduct systems, as a natural generalization of two parameter product systems of Hilbert spaces, introduced by B. Tsirelson in [18]. The Gelfand duality shows that commutative unital C∗-product and subproduct systems are directly related to certain two-parameter families of compact Hausdorff spaces, referred to in this paper as compact super-product systems. Building on this, we define the concept of flattening and show that each compact super-product system can be flattened through a projective limit construction. Furthermore, we are able to define a one-parameter “multiplication” induced by this flattening, which behaves well within the framework of a C∗-product system. Finally, we show that these results hold when considering the appropriate measures for these spaces, as well as the various constructions defined within them.
  • ItemOpen Access
    Role of physiologically relevant hypoxia in neural stem and progenitor cell proliferation, migration and differentiation into oligodendrocytes
    (Faculty of Graduate Studies and Research, University of Regina, 2023-04) Masood, Mahrukh; Buttigieg, Josef; Hart, Mel; Candow, Darren
    Stem cells are undifferentiated cells, defined by their capability to self-renew and differentiate to give rise to different cells of the body. Neural stem and progenitor cells are a type of multipotent stem cell capable of giving rise to the cells of the mature central nervous system (CNS): neurons, astrocytes and oligodendrocytes. The mechanism by which various factors influence stem fate is of wide interest, as these cells play a key role in development, and have a potential role in repair of CNS injury. I investigated the role of physiologically relevant hypoxic levels as a driving force for the proliferation and differentiation of Neural Stem and Progenitor Cells (NSPCs) into Oligodendrocytes (OLs) as well as increased migration. In most research, cells are cultured at 21% O2, which is significantly higher than what these cells, and other cells of the CNS, are typically exposed to. Physioxia is what could be considered low concentrations of O2 in the external environment, but normal in the body. The O2 level in the human body is tightly regulated; particularly low levels of O2 may positively aid in NSPC differentiation through the regulation of certain genes. One mechanism that may aid in the differentiation process is the involvement of transcription factors that are sensitive to changes in O2 levels. Hypoxic Inducible Factor (HIF) is a transcription factor that plays an integral role in the detection of hypoxia and can induce changes in genes responsible for vascular growth (vascular endothelial growth factor (VEGF)), cell migration (matrix metalloproteinase 2 (MMP2)) or A disintegrin and metalloproteinase with thrombosponfin motifs 1 (ADAMTS-1)) and cell differentiation (platelet-derived growth factor (PDGF)). This study demonstrates that a low O2 environment can be confirmed through the upregulation of HIF-1a at low levels of physiologically relevant oxygen levels. Furthermore, the upregulation of VEGF at different O2 concentrations alludes to NSPC proliferation, especially at 5% O2. MMP2 upregulation showed that migration of the OL lineage cells is highest at 5% O2. Lastly, differentiation of NSPCs to OPCs seemed to increase when exposed to low levels of O2 and was the highest at 5% O2. Moderate levels of physiologically relevant oxygen levels such as 5% seem to have the optimal effect on NSPC proliferation, differentiation, and migration as gene expression for several of the gene is highest at that O2 concentration.
  • ItemOpen Access
    Training agents to play cooperative games: A reinforcement learning approach
    (Faculty of Graduate Studies and Research, University of Regina, 2023-07) Marahemi, Sara; Zilles, Sandra; Hamilton, Howard; Smith, Austen
    Due to recent advances in research on reinforcement learning, self-learning agents are capable of accomplishing numerous kinds of tasks in complex environments without any prior knowledge. Recently, deep reinforcement learning algorithms have shown promising results in game-playing tasks that were previously impractical, including playing video games directly from raw screen pixels. In this project, we created a game engine for a card game called “The Mind”, and used reinforcement learning techniques in order to train agents to master this game. The Mind is a multi-player cooperative card game with the challenge of synchronizing the agents’ actions. We used Q-learning and deep Q-learning to estimate a Q-function which describes an agent’s best action to take at any state of the game. In this research, we implemented three types of agents based on two different reinforcement learning algorithms. The results showed that our trained agents performed better than random agent models. The highest testing win rate using the Q-learning algorithm was 86%. We also designed a reinforcement learning strategy, called observer learning, in which an agent updates its knowledge not only based on the feedback to its own actions, but also based on the feedback other agents receive for their actions. We reached the best testing win rates of nearly 99% for two Q-learning agents using our observer learning strategy in four levels of The Mind.
  • ItemOpen Access
    Detection of DoS and DDoS attacks on 5G network slices using deep learning approach
    (Faculty of Graduate Studies and Research, University of Regina, 2023-09) Khan, Md. Sajid; Shahriar, Nashid; Yao, Yiyu; Louafi, Habib; Morgan, Yasser
    A new degree of connectedness and interaction has been introduced by the development of 5G networks. By dividing a physical network into several logical networks, 5G network slicing is a special feature that gives network operators the ability to allocate specific resources and services to various applications and customers. However, 5G network slicing is susceptible to cyberattacks, particularly Denial-of-Service (DoS) or Distributed Denial-of-Service (DDoS) attacks, just like any other network. Such attacks can have a significant negative effect on network performance, degrading services and reducing the availability of slices. The primary objective of this thesis is to examine the impact of DoS/DDoS attacks on 5G network slicing and their potential to disrupt the performance of legitimate users and slice availability. Additionally, a novel dataset specifically tailored to DoS/DDoS attacks in 5G network slicing is generated, as there is no available dataset based on a 5G network slice. Through extensive research, key features relevant to DoS/DDoS attacks are identified and prioritized. To categorize and detect different types of DoS/DDoS attacks, two deep learning techniques, namely the convolutional neural network (CNN) and the Bidirectional Long Short-Term Memory (BLSTM) models, are employed. These models not only utilize the newly created dataset but also enable comparison with existing datasets to assess their effectiveness. This thesis emphasizes how crucial it is to create strong security measures to guard against DoS/DDoS attacks on 5G network slicing. A step in the right direction toward reaching this goal is the construction of a deep learning model for the classification, detection, and production of a new dataset specifically for 5G network slicing. To keep enhancing the security and stability of 5G network slicing, more study in this area will be required. The results indicate that the proposed models have a high accuracy rate of 99.96% in distinguishing different types of DoS/DDoS attacks within the networking slice environment. This achievement is noteworthy as it pertains to a novel context. Additionally, the newly developed models exhibit comparable performance in terms of other confusion metrics. To verify the research outcome, some well-known data sets are used to show the results.
  • ItemOpen Access
    Analyzing the effectiveness of Covid-19 vaccines among different age groups using multinomial logistic regression model
    (Faculty of Graduate Studies and Research, University of Regina, 2023-05) Khalid, Arfa; Deng, Dianliang; Volodin, Andrei; Peng, Wei
    This study is conducted to evaluate the effectiveness of Covid-19 vaccines in different age groups in Saskatchewan, Canada. Data was collected between September 2021 and December 2021, and a statistical method called multinomial logistic regression was used to analyze the relationships between multiple categorical variables. In this study, the categorical variables were the age groups and the vaccination status (fully vaccinated cases, partially vaccinated cases, and unvaccinated cases) of the individuals with the interaction effect of rate of cases. The mathematical proof for the multinomial logistic regression model with interaction effect was derived in this study. The study demonstrated the effectiveness of Covid-19 vaccines among vaccinated age groups and provided theory and practical application of the multinomial logistic regression model. Results show that there is a statistically significant impact of age group and vaccination status on the effectiveness of Covid-19 cases in Saskatchewan. Specifically, there is a difference in vaccine effectiveness based on age groups and vaccination status. The findings of this study provide crucial insights for policymakers and public health officials to optimize vaccination rollout strategies and control the spread of Covid-19. Overall, this study represents an important step in the ongoing efforts to understand the effectiveness of Covid-19 vaccines and to develop policies and interventions that can help mitigate the pandemic impact.
  • ItemOpen Access
    Multiple independent lineups: A procedure for corroborating eyewitness identification evidence in children
    (Faculty of Graduate Studies and Research, University of Regina, 2023-06) Carr, Shaelyn; Kaila, Bruer; Jeff, Loucks; Chris, Oriet; Emily, Pica
    Child eyewitnesses exhibit problematic choosing on police lineups at a higher rate than adults (Fitzgerald & Price, 2015), which is an issue as mistaken eyewitness testimony is a leading cause of wrongful convictions (National Registry of Exonerations, 2019). This study examined a novel eyewitness technique to use with children, the multiple independent lineup (MIL) technique, to assess facial identification accuracy. A total of 486 children (60% male, 39% female, and 1% other; Mage = 8.59) witnessed a live event and, the following day, engaged in a lineup identification task (i.e., single simultaneous face lineup or the multiple independent lineup technique). Largely, the results found support for the multiple independent lineup technique to help infer the accuracy of child eyewitnesses. Interestingly, children of all ages performed similarly on the multiple independent lineup technique. The results also revealed that facial identification responses are similar between the two lineup conditions (i.e., single simultaneous lineup and multiple independent lineup technique). Implications and future directions are discussed.
  • ItemOpen Access
    Efficacy of a brief online mindfulness and self-compassion intervention (Mind-OP+) to increase connectedness: Randomized controlled trial
    (Faculty of Graduate Studies and Research, University of Regina, 2023-07) Bueno, Christine Frances Bast; Beshai, Shadi; Hadjistavropoulos, Heather; Sharpe, Donald; O'Rourke, Hannah
    Connectedness is defined as a connection with others that promotes well-being. Although studies examining connectedness are few to date, the extant literature on closely related concepts suggests connectedness is associated with reduced symptoms of psychological and physical disorders and higher overall well-being. Cultivating feelings of connectedness also appears to encourage prosocial behaviour, such as volunteering or donating to charity. Mindfulness and compassion interventions (MBIs) may be adapted to cultivate feelings of connectedness, thereby unlocking a protective mechanism in mental health and beyond. Further, brief self-guided MBIs are particularly promising, given demonstrations of their efficacy combined with their potential for wide scalability and dissemination. Accordingly, this author examined the effectiveness of an augmented, self-guided, brief, online mindfulness and self-compassion-based intervention (Mind-OP+) to facilitate perceptions of connectedness in undergraduate students. Total of 117 undergraduate students were randomized into a waitlist (n = 55) or Mind-OP+ (n = 62) condition. Participants in the Mind-OP+ condition completed five modules at a pace of one module per week. Correlation analyses with participants that passed baseline attention-checks (n = 101 ) revealed that social connectedness at baseline was correlated positively with mindfulness and self-compassion, and correlated negatively with fears of compassion, depression, anxiety, and stress. Relatedness at baseline was correlated positively with mindfulness and negatively with fears of compassion and depression, stress, and anxiety. Intent-to-treat mixed-model analyses on all randomized participants indicated that, compared to participants in the waitlist condition, participants in Mind-OP+ reported increased feelings of social connectedness (d = 0.81) and relatedness (d = 0.64) at post-treatment, and increased feelings of social connectedness (d = 0.80) and relatedness (d = .38) at one-month follow-up. Mediation analyses completed with protocol adherent participants at post-treatment (n = 47) demonstrated no statistically significant mediation of self-compassion or mindfulness scores on the relationship between group membership and connectedness nor relatedness scores at post-treatment. These findings provide support for the use of brief, accessible, self-guided interventions to cultivate connectedness. Larger, more definitive trials should compare the effects of Mind-OP+ for connectedness against an active control, and examine whether the effects on connectedness are independent of effects of reducing psychological disorder symptoms. This intervention holds promise as an option for those seeking protective factors for their mental health and general resiliency.
  • ItemOpen Access
    Self-training for cyberbully detection: Achieving high accuracy with a balanced multi-class dataset
    (Faculty of Graduate Studies and Research, University of Regina, 2023-08) Ahmadinejad, Mohamad Hosein; Nashid, Shahriar; Lisa, Fan; Samira, Sadaoui; Andrei, Volodin
    Cyberbullying has become an alarming issue in the digital era, causing significant harm to its victims. The development of automated methods for detecting cyberbullying in social media is of paramount importance to safeguard vulnerable individuals. In this thesis, we propose a robust approach based on Machine Learning (ML) and Deep Learning (DL) techniques for cyberbully detection in social media platforms. Our approach involves the meticulous curation of a balanced dataset specifically designed for training the ML/ DL models. To overcome the challenge of limited labeled data, we employ a semi-supervised self-training algorithm, which effectively expands the size of the labeled dataset. By leveraging real-world social media data, we train and test the model, evaluating its performance using key metrics such as precision, recall, and F1-score. In addition, we present our meticulously annotated dataset comprising 99,991 tweets, which we have made publicly available for future scientific investigations. This dataset serves as a valuable resource for further research in this field, facilitating the development and evaluation of novel techniques for cyberbully detection. Our results underscore the near-perfect performance of the proposed approach in the context of cyberbully detection, reaffirming the efficacy of ML and DL techniques for addressing this pervasive problem. These findings offer crucial insights for future research endeavors in this domain and hold practical implications for the development of automated systems capable of detecting and combating cyberbullying in social media platforms. By continuously advancing our understanding of cyberbullying detection and developing sophisticated ML and DL models, we can foster safer digital environments and protect individuals from the detrimental effects of cyberbullying.
  • ItemOpen Access
    Learning to listen differently: Witnessing survivor testimony and implications for ethical responses
    (Faculty of Graduate Studies and Research, University of Regina, 2023-07) Steffany, Salloum; Cappello, Michael; Molina-Giron, Alison; Sterzuk, Andrea; Anderson, Brenda
    During the Educators for Solidarity Initiative in 2017, participants bore witness to survivor testimony about the 1981 massacre that took place on the banks of the Rio Lempa in El Salvador. It is the experience of bearing witness to survivor testimony with participants who accompanied me on this journey, whose affective responses, unsettlement, and understanding of settler responsibilities are at the center of this study. Explored through autoethnography and life writing methodologies this qualitative study aimed to answer the central research question: “Does the act of witnessing survivor testimony at the Rio Lempa impact the obligations of (settler) witnessing and ethical responses? If so, in what ways?” Data analysis of the affective and ethical responses to bearing witness to survivor testimony of four white, settler women revealed that the participants were address-able and response-able to the story/teller. The act of witnessing survivor testimony at the Rio Lempa impacted the obligations of (settler) witnessing and ethical responses. The ethical responses to witnessing survivor testimony involved: acknowledgement, remembrance, recognition that the testimony was of consequence, retestifying, cultivating relationships, affective learning, critical empathy, listening with humility and vulnerability. Witnessing testimony abroad increased these settler participant’s capacity to become more relational towards Indian Residential School survivor testimony as they continue to interrogate their response-ability in relation to Indigenous peoples.
  • ItemOpen Access
    An intelligent system approach for predicting the risk of heart failure
    (Faculty of Graduate Studies and Research, University of Regina, 2023-08) Raihan Khan Rabbi, Imran; Peng, Wei; Mehrandezh, Mehran; Khondoker, Mohammad; Jia, Na
    Heart failure is a chronic, progressive condition in which the heart muscle is unable to pump enough blood to meet the body’s needs for blood and oxygen. It is a severe and long-term condition and there are several complications from heart failure that include irregular heartbeat, sudden cardiac arrest, heart valve problems, pulmonary hypertension, kidney damage, liver damage, malnutrition etc. According to the World Health Organization (WHO), the number one cause of death in cardiovascular diseases (CVD) is estimated at 17.9 million a year, which accounts for 31% of all deaths worldwide. The majority of heart patients are diagnosed at a stage of high risk since the early screening and diagnosis of any heart disease is complicated and the particular medical exams are expensive. The current therapeutic approaches lose their effectiveness at this time, which can have deadly repercussions. To lower the mortality rate, novel methods for the early identification of cardiac disease are therefore vital. The research aims to create intelligent systems that can help doctors identify heart disease more quickly and affordably. The likelihood of a patient's survival will rise with the early discovery of potential damage in the system of the heart. The thesis provides a Fuzzy Inference System approach and Feed Forward Back Propagation Neural Network approach to develop intelligent systems based on some input parameters. There are so many factors that can affect the system of the heart. This research uses eleven major parameters to predict the risk of heart failure. The primary outcome of this study is that modelling based on artificial intelligence approaches is far more successful than what is currently available in the medical field for the early detection of heart disease. The performance of the developed systems has been evaluated by a confusion matrix based on 221 datasets collected from a valid source. The obtained result demonstrates that the performance parameters of the FIS model provide superior results compared to the ANN model. The developed FIS system's accuracy, precision, sensitivity, and specificity are 90.50%, 90.91%, 90.50% and 90.31%, respectively. A Graphical User Interference (GUI) is developed using the MATLAB App designer tool to facilitate the system’s practical applicability for the end-users.
  • ItemOpen Access
    Cellular network KPI prediction on simulated 5G-NR V2N traffic dataset using machine learning
    (Faculty of Graduate Studies and Research, University of Regina, 2023-03) Pusapati, Suryanarayanaraju; Peng, Wei; Khan, Sharfuddin; Maciag, Timothy; Wang, Zhanle
    The arrival of 5G has brought a promise of better connectivity for users, but also a challenge for cellular networks to maintain high-quality service and energy efficiency. To optimize the network and meet user demands, a resource management system is used to allocate resources in the 5G Radio Access Network (RAN). However, manual tuning of this system is complex and time-consuming. By predicting the future behavior of Network Key Performance Indexes (KPIs) of the 5G network using Artificial Intelligence (AI) and its subfield, Machine Learning (ML), this study can automate the operations of the Resource Management system, improve resource allocation, and satisfy QoS requirements while optimizing energy consumption. However, to develop a better performing ML model, a high-quality dataset is essential. Since there is a lack of open datasets available on 5G systems, many researchers rely on synthetically generated datasets. This thesis work utilized 5G simulation tool to simulate 5G New Radio (NR) Vehicle-to-Network (V2N) communication using OMNeT++ and SUMO simulators. The NR V2N communication was simulated in a Regina downtown scenario using the proposed simulation framework, and the simulation results were processed using the developed new_df Python module into synthetic datasets that were validated by comparing with technical specifications to ensure their quality. The synthetic datasets were then used to develop proposed Network KPI prediction models using ML. Three ML models are trained and tested, which can predict multiple KPIs, bi-directional Signal to Interference and Noise Ratio (SINR) and classify uplink Channel Quality Indicator (CQI) respectively. The multi-output regression models have shown outstanding performance with MSE as low as 0.002, and the multi-class classification model has a high accuracy. In summary, this study contributes to the development of efficient and automated Resource Management systems for 5G networks using AI and ML techniques. An open source V2N simulation framework was developed using OMNeT++ and SUMO simulators that can simulate 5G-NR V2N communication in a realistic urban scenario. Moreover, a new_df Python function was developed for processing simulation results into an aggregated dataset and spatiotemporal dataset, providing a high-quality dataset that can be used to train and test ML models for predicting Network KPIs of the 5G network.
  • ItemOpen Access
    Efficient coverage path planning and navigation of mobile farming robots
    (Faculty of Graduate Studies and Research, University of Regina, 2023-05) Nasirian, Behnam; Mehrandezh, Mehran; Janabi-Sharifi, Farrokh; Stilling, Denise; Wang, Zhanle
    Autonomous farming uses technologies such as robotics and artificial intelligence to automate agricultural operations, reduce labor requirements, and improve productivity. With a shortage of skilled labor in many parts of the world and increasing demand for food due to population growth, autonomous farming can help farmers increase efficiency, manage larger farms, and improve crop yields and profitability. Furthermore, autonomous farming technologies can lead to improved environmental sustainability by reducing the use of pesticides and fertilizers through targeted and precise application. Mobile robots, such as self-driving tractors, are increasingly being used to automate various agricultural operations such as planting, spraying, and harvesting using advanced algorithms and sensors. The navigation stack, consisting of a set of algorithms and software tools for mobile robot navigation and path planning, is a critical component of autonomous farming. This work presents algorithms for coverage path planning, line-following control, and local navigation used in the navigation stack of a mobile robot, enabling it to operate safely and accurately, thus increasing efficiency and decreasing costs in farming operations. Coverage path planning (CPP) is an essential component of autonomous farming that enhances the efficiency of agricultural tasks. It provides benefits in terms of cost, time, and quality of coverage. Coverage path planning refers to the process of generating an uninterrupted, collision-free path for a mobile robot to cover a designated area of interest. The aim of this work is to reduce the time and cost of agricultural operations by creating an optimal coverage path using a graph-based representation of the field. The proposed technique is compared to other existing techniques, and it is demonstrated that the proposed technique generates a coverage path with shorter traveled distance (less overlapping) and fewer turns. The mobile robot should be capable of following the coverage path generated for a field. The coverage path in agricultural applications includes lots of back-and-forth straight-line motions. Therefore, an efficient control approach is required to achieve and follow desired straight lines on the coverage path precisely (to avoid overlaps and skipped areas) and in a time and cost efficient manner. A time-varying model predictive control (MPC) has been designed to determine the appropriate steering angle of a mobile robot for line-following control. To evaluate the performance and efficiency of the designed time-varying MPC technique, it has been compared with a proportional control technique as a common control method for line-following problems. The time-varying MPC proves to outperform the proportional controller in terms of performance and cost. The coverage path in a field is a global path, which is generated based on a prior knowledge of the field and obstacles. However, if anything changes in this a priori known map of the field, motion plans need to change accordingly, which brings out the notion of local motion planning and navigation around obstacles. In this work, a human-analogous learning technique has been proposed that can learn from a human operator trained in a simulated environment under a learning-by-doing paradigm. A human-in-the-loop simulator, utilized as the training environment for sensor-based motion planning, is developed. A neurofuzzy- based steering algorithm is derived from data collected from a trained human operator. The simulation results are compared to that cited in literature. The proposed algorithm generates superior steering without the need to setting up a cost function and tuning its parameters to generate an efficient local path.