University of Regina Institutional Repository
Not a member yet
15455 research outputs found
Sort by
Virtual Methodology for Household Waste Characterization During The Pandemic in An Urban District of Peru: Citizen Science for Waste Management
No full textThis study examines urban plastic waste generation using a citizen science approach in six Latin American countries during a global pandemic. The objectives are to quantify generation rates of masks, gloves, face shields, and plastic bags in urban households using online survey, and perform a systematic cross-jurisdiction comparisons in these Latin American countries. The per capita total mask generation rates ranged from 0.179 maskcap-1day-1 to 0.915 maskcap-1day-1. A negative correlation between the use of gloves and masks is observed. Using the average values, the approximate proportion of masks, gloves, shields, single-use plastic bags was 34:5:1:84. We found that most studies overestimated face mask disposal rate in Latin America due to the simplifying assumptions on number of masks discarded per person, masking prevalence rate, and average mask weight. Unlike other studies, end-of-life PPE quantities were directly counted and reported by the survey participants. Both of the conventional weight-based estimates and the proposed participatory survey are recommended in quantifying COVID waste. Participants' perception based on the Likert scale is generally consistent with the waste amount generated. Waste policy and regulation appear to be important in daily waste generation rate. The results highlight the importance of using measured data in waste estimates.The research reported in this paper was supported by a grant from the Natural Sciences and Engineering Research Council of Canada (RGPIN-2019-06154) to the corresponding author, using computing equipment funded by FEROF at the University of Regina.This version of the article has been accepted for publication, after peer review (when applicable) and is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: https://doi.org/10.1007/s00267-022-01610-1
Machine learning enabled wireless sensor network for partial discharge detection
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Applied Science in Electronic Systems Engineering, University of Regina. xv, 84 p.Partial Discharge (PD) is one of the leading causes of failure in high voltage electrical equipment, making it an important metric for predictive and preventative maintenance. Early detection of PD and corresponding maintenance can prevent catastrophic failures and improve safety and reliability. PD occurs when insulation between conductors is partially bridged causing electrical stress on the surface of the insulation. The PD phenomenon typically results in pulses of duration 1 μs or less. PD can be monitored using Off-line and On-line methods. While on-line is a more attractive option as it does not require system downtime. Some On-line methods include electrical detection, acoustic detection, and chemical detection. Sensors used for on-line PD monitoring include but are not limited to: Ultra High Frequency (UHF) Sensors, High Frequency Current Transformers (HFCT), Accoustic Emmissions (AE) sensors, Transient Earth Voltage (TEV) Sensors. PD signals are not easily recognizable in the raw sensor data due to the abundance of noise from the environment in which the equipment resides. PD monitoring also results in a large data due to the sampling rate required to catch the high frequency PD pulses. This leads to challenges with de-noising, identifying and processing PD data. PD has been widely studied and detection systems are currently available. These challenges have been identified in existing literature and solutions have been proposed and examined. However, existing research is limited in the use of Wireless Sensor Networks (WSN) for PD detection.
This thesis provides an end-to-end system for PD detection using a Machine Learning (ML) Enabled WSN. This system is implemented using signal condition techniques which include filtering, maximum pooling, and the Discrete Wavelet Transform (DWT). The ML algorithms, Support Vector Machine (SVM) and Convolutional Neural Network (CNN) are implemented and tested with PD data and their effectiveness is evaluated. Finally, a WSN is implemented and the signal conditioning and ML algorithm is configured on end-devices which transmit the PD data to a centralized base station. With the implementation of the proposed system, the use of a WSN for PD detection becomes a much more viable option as the data required to be transmitted is decreased significantly. We are able to decrease the required transmission rate from 960 Mbps when transmitting raw PD sensor data to 60.4 kbps when transmitting the conditioned data.Studentye
Release: Backgrounder: University of Regina researchers receive Social Sciences and Humanities Research Council (SSRHC) Insight Development Grants (IDG) worth $109,107
No full textStaffn
Shifting focus: Feasibility of online mindfulness meditations as an adjunct to tailored internet-delivered cognitive behaviour therapy
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Science in Clinical Psychology, University of Regina. x, 117 p.Background. Public safety personnel (PSP) are at an increased risk for developing mental disorders compared to the general population. The PSP Wellbeing Course is a transdiagnostic internet-delivered cognitive behaviour therapy (ICBT) course tailored to assist PSP to manage symptoms of depression, anxiety, and posttraumatic stress injuries using cognitive behavioural strategies. While the effectiveness of this course is supported by evidence (Hadjistavropoulos, McCall, et al., 2021), incorporating mindfulness as an additional strategy to assist PSP with symptoms could potentially improve the program. Mindfulness interventions can help people learn to experience the world and their reactions to the world in open and non-judgmental ways, which may complement the existing PSP Wellbeing Course content. Objective. To examine the feasibility of mindfulness meditations in the PSP Wellbeing Course. Methods. The current study used a mixed-methods design including quantitative and qualitative data collection. Participants included 40 treatment-seeking PSP who were asked to complete five mindfulness meditations including grounding, loving kindness, awareness of breath, awareness of five senses, and body scan meditations alongside five core PSP Wellbeing Course lessons. On a weekly basis, participants indicated how often they participated in mindfulness meditation. Participants completed measures (i.e., anger, depression, anxiety, posttraumatic stress disorder, insomnia, and resilience) pre- and post-treatment, and treatment satisfaction scales post-treatment. There were 12 participants who also completed an interview about perceptions of the mindfulness meditations. Results. There were 27 (67.5%) participants who reported using the mindfulness meditations, putting in 4.8 minutes (SD = 8.1) of practice each week. The course was associated with significant improvements in the primary symptom measures, functional impairment, and resilience.
Practice was not associated with improved outcomes. Most interviewed participants described the mindfulness meditations as beneficial, helping to slow down and regulate their bodies and emotions. Participants also reported challenges with the meditations, such as discomfort sitting with their feelings. Participants provided suggestions for improvement (e.g., creating shorter meditations, adding clear signals to indicate the end of meditations). Discussion. The current study demonstrates the feasibility of adding mindfulness meditations to the PSP Wellbeing Course with many PSP making use of and reporting benefits from the meditations. Nevertheless, improvements could be made to improve use of meditations. Future research appears warranted to systematically test the benefits of adding mindfulness to the PSP Wellbeing Course as well as longer term outcomes of the meditations.Studentye
Public acceptance of facial recognition technology: Surveying attitudes, preferences, and concerns to inform policy development
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Public Policy, University of Regina. ix, 89 p.One application of artificial intelligence (AI) in public sector governance is facial recognition technology (FRT), which is used to identify or discover individuals by comparing an image of their face to a database of known faces for a match. Along with these applications, however, concerns surrounding FRT use by governments have emerged as critics raise issues not only about the technology itself, but also the implications for the expansion of the ‘surveillance society’ and specific concerns such as demonstrated racial biases in FRT. While FRT continues to be developed and used, and governments struggle to develop a legislative and regulatory response, the question of public acceptance of FRT emerges as timely and important for ongoing policy deliberations. This thesis reports on an empirical test of what the public deems acceptable in the context of FRT applications used in the public sector, focusing on citizens’ attitudes towards, preferences for, and concerns about public sector use of FRT. A survey of 266 residents of two comparative provincial jurisdictions in Canada — Saskatchewan and Ontario — gathered information on attitudes towards FRT used in a variety of settings using nine hypothetical scenario vignettes of FRT use. Descriptive statistics, correlation statistics, and regression analysis is used to identify which socio-economic and demographic factors predict support for public sector use of FRT. The findings from this research have implications for the adoption of FRT in the public sector and for the development of legislation and regulation in response to its use. Results indicate safety and security as a public priority and a general overall support for FRT use by public sector agencies regardless of sociodemographic characteristics, with the least favourable use being general public surveillance. Citizen service uses (particularly airport security) yielded the highest levels of support. The use of FRT by public sector agencies will require overt purpose and perceived public value. Public policy that has a strong focus on personal privacy is needed to balance the interests of Canadians with government public service.Studentye
Feature Story: Culture Days: Artists and biologists find inspiration in the University of Regina's George F. Ledingham Herbarium
No full textDeep in the halls of the University lives a curiosity cabinet of more than 70,000 specimens of plants, flowers, mosses, and lichens. The George F. Ledingham Herbarium is the legacy of its founder, whose life's passion was collecting plants and preserving the natural prairie landscape. Some specimens in the collection date back as far as the 1920s and range into the 2000s. The herbarium, which is maintained by the Department of Biology is of special use to taxonomists, field ecologists, and those working with endangered plant species.Staffn
Release: Creator Wayne Goodwill has presented a traditional Buffalo Winter Count Robe to the University of Regina
No full textStaffn
Clustering and dimensionality reduction for time-series service monitoring data
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Computer Science, University of Regina. xx, 193 p.Service monitoring applications allow customers to measure the performance, availability, and resolve application issues before they affect users. Since service monitoring applications continuously produce data to monitor their availability, therefore, high dimensionality, unlabeled data and changing data distribution are all prevalent. In this thesis, we efficiently address these three issues using the constructed service monitoring dataset. Higher dimensionality means higher computational cost to perform training and often leads to over-fitting while l earning a model. Furthermore, in the presence of high dimensionality, data are highly correlated resulting in insignificant and irrelevant features. These features have less impact on the prediction. To this end, the first part of the thesis conceptually and empirically explores the most representative dimensionality reduction (DR) methods from different categories.
Next, we construct a new and challenging End-to-End (E2E) service monitoring dataset by extracting heterogeneous sub-datasets from multiple subservers, tackling data incompleteness in each sub-dataset using several imputation techniques, and fusing all the optimally imputed sub-datasets. This target dataset is highly dimensional, temporal, unlabeled, and non-linear. As the dataset is new, based on robust clustering approaches, we thoroughly assess the quality of the initial dataset and the reconstructed datasets (same dimensionality as the initial dataset) produced with Deep and Convolutional AutoEncoders. The experiments disclose that the reconstructed dataset with Deep AutoEncoder is the most performing.
Later, we propose an ensemble-based DR approach to effectively handle the high-dimensionality of the E2E dataset. The approach combines Deep AutoEncoder with Kernel Principal Component Analysis to produce better data, and then reduce the feature space respectively. Due to the massive size of the dataset, we divide it into six weekly sub-datasets. We show that no vital information is lost for the reduced sub-datasets using the reconstruction error and total explained variance ratio. Based on time-series data clustering methods, and metrics, we thoroughly evaluate the efficacy of the ensemble approach. As the initial dataset is unlabeled (so are the reduced sub-datasets), we improve the previously developed ensemble-based DR approach by further combining it with incremental DR to improve clusters’ performance and increase the cluster labels’ confidence. We consider the weekly datasets as chunks for the experimental purpose. The experiments reveal that clustering performances increase significantly after utilizing the improved ensemble-based DR. Hence, the clusters’ labels are considered as the target class labels.
Finally, to process the incoming data for any service monitoring application, it is critical to classify data accurately in real-time. Hence, we consider the labeled data chunks as incoming data, and propose an adaptive classification framework using Learn++ that also handles evolving data distributions. This approach sequentially predicts and updates the monitoring model with new data, and gradually forgets past knowledge. We employ consecutive data chunks to evaluate the performance of the predictors incrementally. The experimental results demonstrate that the proposed method provides high detection rates and low misclassification rate for most of the adaptive chunks.Studentye
Machine learning-based intrusion detection system in advanced metering infrastructure
No full textA Thesis Submitted to the Faculty of Graduate Studies and Research In Partial Fulfillment of the Requirements for the Degree of Master of Applied Science in Electronic Systems Engineering, University of Regina. xi, 94 p.Smart meters are quickly being introduced to households all over the world. Consumers will benefit significantly from Advanced Metering Infrastructure (AMI). This high-tech equipment, however, is vulnerable to serious threats. To detect attacks, many authors recommend using Intrusion Detection Systems (IDSs). An IDS use a classifier to distinguish between known and unknown threats. It is challenging to choose the proper IDS category, however, as no single category can accurately identify all types of attacks. A smart grid's autonomous meter reading system is built on Advanced Metering Infrastructure (AMI). It blends traditional utility operations and asset management methods with wide-ranging elements of machinery, such as (a) automated metering; (b) communication networks; and (c) data management systems, to enhance the grid's interface with customers and utility providers. AMI enables smart meters and utilities to communicate in a two-way mode, regarding data such as power usage, price, upgraded firmware, remote disconnection, issue or outage detection, and exclusion notifications. One of the most significant obstacles to AMI's global acceptance is its security. We investigated the topic of detecting malicious assaults in AMI in this thesis. By the time the national Smart meter goes live, cybersecurity experts predict that four major categories of attacks will be prevalent. Data attacks affect the smart meter to cause erroneous decisions/actions, by attempting to adversely inject, change, or remove data or control commands in the networking ow. By examining energy use statistics, a privacy attack attempts to learn or infer customers' confidential information. Smart meters record power to use data numerous times each hour, and the precise data may quickly disclose consumers' private, physical activity. A network availability attack can take the form of a denial of service (DoS) attack, which will cause a delay or loss in data connection. This thesis emphasis on attacks that threaten the availability and integrity of the entire system. We describe a hierarchical, distributed approach for Smart meters, which merges the feature engineering of preprocessing with machine learning classifiers, and ne tunes the hyper-parameters, using voting classifiers to improve detection accuracy and processing time. Different types of classifiers, decision trees (DT), random forests (RF), and Naive Bayes (NB) classifiers tend to be used in this thesis. AWS Sage-Maker, AWS Sage-Maker Autopilot, and Google Colab were used to test the overall performance of the system. This thesis contributes to existing research by demonstrating how much AMI attack detection enhancing strategies, and different types of IDS techniques in AMI networks. According to analyze the output, three combined classifiers outperformed single classifiers in terms of accuracy and processing time.Studentye
Within-Person Variability Contributes to More Durable Learning of Faces
No full text© Copyright 2022 Canadian Psychological Association. https://doi.org/10.1037/cep0000282. This article may not exactly replicate the final version published in the CPA journal. It is not the copy of record.Exposure to the natural, unsystematic within-person variability present across different encounters with a face (e.g., differences in emotion, make-up, and hairstyle) increases the likelihood the face will be recognized despite changes in appearance. In most studies, participants’ memories are tested with a matching task administered shortly after exposure to a set of training images. In the real world, however, the time between when a face is first encountered and when it needs to be identified can be much longer. We hypothesized that in addition to facilitating acquisition of a representation of a face, unsystematic variability might also lead to better retention. To test this, in two experiments participants were randomly assigned to one of three training conditions: a) no variability (still image), b) systematic variability (changes in camera angle and pose in an otherwise constant setting), and c) unsystematic variability (changes in hairstyle, makeup, clothing, and setting). Participants completed a sorting task 15 minutes and 5 days after viewing the target identity. Unsystematic variability led to better recognition than systematic variability, and this benefit was not reduced after a 5 day delay. Although participants expected their memory to be worse with a 5 day delay than with a 15 minute delay, both overall accuracy and the advantage for training with unsystematic variability were virtually unaffected. The results suggest that exposure to unsystematic variability influences not only the initial acquisition of faces, but also contributes to establishing a durable, flexible representation of faces in memory.This research was supported by an NSERC Discovery Grant RGPIN-2017-06005 to C.O.Facultyye