Computer Engineering Commons^™

Investigation Into The Perceptually Informed Data For Environmental Sound Recognition, Chenglin Kang

Dissertations

Environmental sound is rich source of information that can be used to infer contexts. With the rise in ubiquitous computing, the desire of environmental sound recognition is rapidly growing. Primarily, the research aims to recognize the environmental sound using the perceptually informed data. The initial study is concentrated on understanding the current state-of-the-art techniques in environmental sound recognition. Then those researches are evaluated by a critical review of the literature. This study extracts three sets of features: Mel Frequency Cepstral Coefficients, Mel-spectrogram and sound texture statistics. Two kinds machine learning algorithms are cooperated with appropriate sound features. The models are …

Comparing Defeasible Argumentation And Non-Monotonic Fuzzy Reasoning Methods For A Computational Trust Problem With Wikipedia, Ryan Kirwan

Dissertations

Computational trust is an ever-more present issue with the surge in autonomous agent development. Represented as a defeasible phenomenon, problems associated with computational trust may be solved by the appropriate reasoning methods. This paper compares two types of such methods, Defeasible Argumentation and Non-Monotonic Fuzzy Logic to assess which is more effective at solving a computational trust problem centred around Wikipedia editors. Through the application of these methods with real-data and a set of knowledge-bases, it was found that the Fuzzy Logic approach was statistically significantly better than the Argumentation approach in its inferential capacity.

Predicting Violent Crime Reports From Geospatial And Temporal Attributes Of Us 911 Emergency Call Data, Vincent Corcoran

Dissertations

The aim of this study is to create a model to predict which 911 calls will result in crime reports of a violent nature. Such a prediction model could be used by the police to prioritise calls which are most likely to lead to violent crime reports. The model will use geospatial and temporal attributes of the call to predict whether a crime report will be generated. To create this model, a dataset of characteristics relating to the neighbourhood where the 911 call originated will be created and combined with characteristics related to the time of the 911 call. Geospatial …

Enhancing Partially Labelled Data: Self Learning And Word Vectors In Natural Language Processing, Eamon Mcentee

Dissertations

There has been an explosion in unstructured text data in recent years with services like Twitter, Facebook and WhatsApp helping drive this growth. Many of these companies are facing pressure to monitor the content on their platforms and as such Natural Language Processing (NLP) techniques are more important than ever. There are many applications of NLP ranging from spam filtering, sentiment analysis of social media, automatic text summarisation and document classification.

Performance Comparison Of Hybrid Cnn-Svm And Cnn-Xgboost Models In Concrete Crack Detection, Sahana Thiyagarajan

Dissertations

Detection of cracks mainly has been a sort of essential step in visual inspection involved in construction engineering as it is the commonly used building material and cracks in them is an early sign of de-basement. It is hard to find cracks by a visual check for the massive structures. So, the development of crack detecting systems generally has been a critical issue. The utilization of contextual image processing in crack detection is constrained, as image data usually taken under real-world situations vary widely and also includes the complex modelling of cracks and the extraction of handcrafted features. Therefore the …

Is There A Correlation Between Wikidata Revisions And Trending Hashtags On Twitter?, Paula Dooley

Dissertations

Twitter is a microblogging application used by its members to interact and stay socially connected by sharing instant messages called tweets that are up to 280 characters long. Within these tweets, users can add hashtags to relate the message to a topic that is shared among users. Wikidata is a central knowledge base of information relying on its members and machines bots to keeping its content up to date. The data is stored in a highly structured format with the added SPARQL protocol and RDF Query Language (SPARQL) endpoint to allow users to query its knowledge base.

Forecasting Anomalous Events And Performance Correlation Analysis In Event Data, Sonya Leech [Thesis]

Dissertations

Classical and Deep Learning methods are quite common approaches for anomaly detection. Extensive research has been conducted on single point anomalies. Collective anomalies that occur over a set of two or more durations are less likely to happen by chance than that of a single point anomaly. Being able to observe and predict these anomalous events may reduce the risk of a server’s performance. This paper presents a comparative analysis into time-series forecasting of collective anomalous events using two procedures. One is a classical SARIMA model and the other is a deep learning Long-Short Term Memory (LSTM) model. It then …

Detection Of Offensive Youtube Comments, A Performance Comparison Of Deep Learning Approaches, Priyam Bansal

Dissertations

Social media data is open, free and available in massive quantities. However, there is a significant limitation in making sense of this data because of its high volume, variety, uncertain veracity, velocity, value and variability. This work provides a comprehensive framework of text processing and analysis performed on YouTube comments having offensive and non-offensive contents.

YouTube is a platform where every age group of people logs in and finds the type of content that most appeals to them. Apart from this, a massive increase in the use of offensive language has been apparent. As there are massive volume of new …

An Evaluation Of The Information Security Awareness Of University Students, Alan Pike

Dissertations

Between January 2017 and March 2018, it is estimated that more than 1.9 billion personal and sensitive data records were compromised online. The average cost of a data breach in 2018 was reported to be in the region of US$3.62 million. These figures alone highlight the need for computer users to have a high level of information security awareness (ISA). This research was conducted to establish the ISA of students in a university. There were three aspects to this piece of research. The first was to review and analyse the security habits of students in terms of their own personal …

High-Speed Data Communications For Vehicular Networks Using Free-Space Optical Communications, Yagiz Kaymak

Dissertations

The demand for high-speed Internet access for vehicles, such as high-speed trains (HSTs) and cars, is on the rise. Several Internet access technologies that use radio frequency are being considered for vehicular networking. Radio-frequency communications technologies cannot provide high data rates due to interference, bandwidth limitations, and the inherent limited data rates of radio technology. Free-space optical communications (FSOC) is an alternative approach and a line-of-sight (LOS) technology that uses modulated light to transfer data between two free-space optical (FSO) transceivers. FSOC systems for vehicular networks are expected to provide data rates in the range of Gbps for stationary and …

Computational Intelligence In Steganography: Adaptive Image Watermarking, Xin Zhong

Dissertations

Digital image watermarking, as an extension of traditional steganography, refers to the process of hiding certain messages into cover images. The transport image, called marked-image or stego-image, conveys the hidden messages while appears visibly similar to the cover-image. Therefore, image watermarking enables various applications such as copyright protection and covert communication. In a watermarking scheme, fidelity, capacity and robustness are considered as crucial factors, where fidelity measures the similarity between the cover- and marked-images, capacity measures the maximum amount of watermark that can be embedded, and robustness concerns the watermark extraction under attacks on the marked-image. Watermarking techniques are often …

Analysing Online User Activity To Implicitly Infer The Mental Workload Of Web-Based Tasks Using Defeasible Reasoning, Paul Mara

Dissertations

Mental workload can be considered the amount of cognitive load or effort used over time to complete a task in a complex system. Determining the limits of mental workload can assist in optimising designs and identify if user performance is affected by that design. Mental workload has also been presented as a defeasible concept, where one reason can defeat another and a 5-layer schema to represent domain knowledge to infer mental workload using defeasible reasoning has compared favourably to state-of-the-art inference techniques. Other previous work investigated using records of user activity for measuring mental workload at scale using web-based tasks …

Design, Analysis And Optimization Of Visible Light Communications Based Indoor Access Systems For Mobile And Internet Of Things Applications, Sihua Shao

Dissertations

Demands for indoor broadband wireless access services are expected to outstrip the spectrum capacity in the near-term "spectrum crunch". Deploying additional femtocells to address "spectrum crunch" is cost-inefficient due to the "backhaul challenge" and the exorbitant system maintenance. According to an Alcatel-Lucent report, most mobile Internet access traffic happens indoors. To alleviate the "spectrum crunch" and the "backhaul challenge" problems, visible light communication (VLC) emerges as an attractive candidate for indoor wireless access in the 5G architecture. In particular, VLC utilizes LED or fluorescent lamps to send out imperceptible flickering light that can be captured by a smart phone camera …

Spoilage Detection In Raspberry Fruit Based On Spectral Imaging Using Convolutional Neural Networks, Karthik Kuchangi Jothi Prakash

Dissertations

Effective spoilage detection of perishable food items like fruits and vegetables is essential for retailers who stock and sell large quantities of these items. This research is aimed at developing a non-destructive, rapid and accurate method which is based on Spectral Imaging (SI) used in tandem with Convolutional Neural Network (CNN) to predict whether the fruit is fresh or rotten. The study also aims to determine the number of days before which the fruit rots. This research employs a primary, quantitative and inductive methods to investigate the Deep Learning based approach to detect fruit spoilage. Raspberry fruit in particular has …

A Comparison Of Real Time Stream Processing Frameworks, Jonathan Curtis

Dissertations

The need to process the ever-expanding volumes of information being generated daily in the modern world is driving radical changes in traditional data analysis techniques. As a result of this, a number of open source tools for handling real-time data streams has become available in recent years. Four, in particular, have gained significant traction: Apache Flink, Apache Samza, Apache Spark and Apache Storm. Despite the rising popularity of these frameworks, however, there are few studies that analyse their performance in terms of important metrics, such as throughput and latency. This study aims to correct this, by running several benchmarks against …

Comparing The Effectiveness Of Different Classification Techniques In Predicting Dns Tunnels, Patrick Walsh

Dissertations

DNS is one of the most widely used protocols on the internet and is used in the translation of domain names into IP address in order to correctly route messages between computers. It presents an attractive attack vector for criminals as the service is not as closely monitored by security experts as other protocols such as HTTP or FTP. Its use as a covert means of communication has increased with the availability of tools that allow for the creation of DNS tunnels using the protocol. One of the primary motivations for using DNS tunnels is the illegal extraction of information …

Application Of Synthetic Informative Minority Over-Sampling (Simo) Algorithm Leveraging Support Vector Machine (Svm) On Small Datasets With Class Imbalance, Akshatha Fakkeriah Kallappanamatt

Dissertations

Developing predictive models for classification problems considering imbalanced datasets is one of the basic difficulties in data mining and decision-analytics. A classifier’s performance will decline dramatically when applied to an imbalanced dataset. Standard classifiers such as logistic regression, Support Vector Machine (SVM) are appropriate for balanced training sets whereas provides suboptimal classification results when used on unbalanced dataset. Performance metric with prediction accuracy encourages a bias towards the majority class, while the rare instances remain unknown though the model contributes a high overall precision. There are chances where minority instances might be treated as noise and vice versa. (Haixiang et …

An Application Of Natural Language Processing For Triangulation Of Cognitive Load Assessments In Third Level Education, Luis Alfredo Contreras

Dissertations

Work has been done to measure Mental Workload based on applications mainly related to ergonomics, human factors, and Machine Learning. The influence of Machine Learning is a reflection of an increased use of new technologies applied to areas conventionally dominated by theoretical approaches. However, collaboration between MWL and Natural Language Processing techniques seems to happen rarely. In this sense, the objective of this research is to make use of Natural Languages Processing techniques to contribute to the analysis of the relationship between Mental Workload subjective measures and Relative Frequency Ratios of keywords gathered during pre-tasks and post-tasks of MWL activities …

Can Threshold-Based Sensor Alerts Be Analysed To Detect Faults In A District Heating Network?, Liam Cantwell

Dissertations

Older IoT “smart sensors” create system alerts from threshold rules on reading values. These simple thresholds are not very flexible to changes in the network. Due to the large number of false positives generated, these alerts are often ignored by network operators. Current state-of-the-art analytical models typically create alerts using raw sensor readings as the primary input. However, as greater numbers of sensors are being deployed, the growth in the number of readings that must be processed becomes problematic. The number of analytic models deployed to each of these systems is also increasing as analysis is broadened. This study aims …

Elasticity Measurement In Caas Environments - Extending The Existing Bungee Elasticity Benchmark To Aws's Elastic Container Service, Nora Limbourg

Dissertations

Rapid elasticity and automatic scaling are core concepts of most current cloud computing systems. Elasticity describes how well and how fast cloud systems adapt to increases and decreases in workload. In parallel, software architectures are moving towards employing containerised microservices running on systems managed by container orchestration platforms. Cloud users who employ such container-based systems may want to compare the elasticity of different systems or system settings to ensure rapid elasticity and maintain service level objectives while avoiding over-provisioning. Previous research has established a variety of metrics to measure elasticity. Some existing benchmark tools are designed to measure elasticity in …

Use Of Hyperspectral Images (Hsi) And Convolutional Neural Network (Cnn) To Identify Normal, Precancerous And Cancerous Tissues, Pallavi Jain

Dissertations

Cancer detection has been a great topic of research for a long time, as early detection of cancer can help in increasing the survival rate of patients by providing on time better treatment. A robust system is required in order to detect early-stage cancer as its difficult to identify early-stage cancer from the normal clinical process. The computer vision techniques provide a new way to understand the challenges related to the medical image analysis. This thesis presents the medical image analysis using a combination of Convolutional Neural Network and Hyperspectral Images of cancer patient's tissues. The idea behind choosing the …

From Business Understanding To Deployment: An Application Of Machine Learning Algorithms To Forecast Customer Visits Per Hour To A Fast-Casual Restaurant In Dublin, Odunayo David Adedeji

Dissertations

This research project identifies the significant factors that affects the number of customer visits to a fast-casual restaurant every hour and proceeds to develop several machine learning models to forecast customer visits. The core value proposition of fast-casual restaurants is quality food delivered at speed which means they have to prepare meals in advance of customers visit but the problem with this approach is in forecasting future demand, under estimating demand could lead to inadequate meal preparation which would leave customers unsatisfied while over estimation of demand could lead to wastage especially with restaurants having to comply with food safety …

Can Machine Learning Beat Physics At Modeling Car Crashes?, Gavin Byrne

Dissertations

This study aimed to look at a traditional method used for measuring the severity and principle direction of force of a car crash and see if it could be improved on using machine learning models. The data used was publicly available from the NHTSA database and included descriptions of the vehicle, test and sensors as well as the accelerometer data over the period of the crashes. The models built were SVM classifiers and multinomial regression models. Although the SVM and Regression models were built successfully and gave higher levels of accuracy than the momentum models in terms of the severity, …

Supervised Learning Models To Predict Stock Direction Within Different Sectors In A Bull And Bear Market, Tiffany Razy

Dissertations

Forecasting stock market price movement is a well researched and an alluring topic within the machine learning and financial realm. Supervised machine learning algorithms such as Random Forest (RF) and Support Vector Machines (SVM) have been used independently to gain insight on the market. With such volatility in the market the scope of this study will utilized the RF and SVM in a very volatility market to determine if these models will perform at a high level or outperform each other in both markets. This relative study is performed on 16 stocks in 4 different sectors over the bear market …

Identifying Expert Investors On Financial Microblog Via Artificial Neural Networks, Pierluca Del Buono

Dissertations

In the recent years, thanks to social media platform, a plethora of information has been available to financial investors, that were traditionally dependent from financial institutions advisors. Strategies are now shared among web users, performances of stocks are commented in web communities and hints and suggestions are travelling on the internet with a fast pace, in a way that was unthinkable few years before. Several attempts have been made in the recent past, to predict Market movements and trends from activity of Financial Social Networks participants, and to evaluate if contributions from individuals with high level of expertise distinguish themselves …

Investigating The Application Of Deep Convolutional Neural Networks In Semi-Supervised Video Object Segmentation, Jayadeep Sasikumar

Dissertations

This thesis investigates the different approaches to video object segmentation and the current state-of-the-art in the discipline, focusing on the different deep learning techniques used to solve the problem. The primary contribution of the thesis is the investigation of usefulness of Exponential Linear Units as activation functions for deep convolutional neural architectures trained to perform object semi-supervised segmentation in videos. Mask R-CNN was chosen as the base convolutional neural architecture, with the view of extending the image segmentation algorithm to videos. Two models were created, one with Rectified Linear Units and the other with Exponential Linear Units as the respective …

Beef Cattle Instance Segmentation Using Mask R-Convolutional Neural Network, Mohammad Danish

Dissertations

Maintaining the cattle farm along with the wellbeing of every heifer has been the major concern in dairy farm. A robust system is required which can tackle the problem of continuous monitoring of cows. the computer vision techniques provide a new way to understand the challenges related to the identification and welfare of the cows. This paper presents a state-of-art instance segmentation mask RCNN algorithm to train and build a model on a very challenging cow dataset that is captured during the winter season. The dataset poses many challenges such as overlapping of cows, partial occlusion, similarity between cows and …

The Use Of Persistent Explorer Artificial Ants To Solve The Car Sequencing Problem, Kieran O'Sullivan

Dissertations

Ant Colony Optimisation is a widely researched meta-heuristic which uses the behaviour and pheromone laying activities of foraging ants to find paths through graphs. Since the early 1990’s this approach has been applied to problems such as the Travelling Salesman Problem, Quadratic Assignment Problem and Car Sequencing Problem to name a few. The ACO is not without its problems it tends to find good local optima and not good global optima. To solve this problem modifications have been made to the original ACO such as the Max Min ant system. Other solutions involve combining it with Evolutionary Algorithms to improve …

Can Deep Learning Techniques Improve The Risk Adjusted Returns From Enhanced Indexing Investment Strategies, Anthony Grace

Dissertations

Deep learning techniques have been widely applied in the field of stock market prediction particularly with respect to the implementation of active trading strategies. However, the area of portfolio management and passive portfolio management in particular has been much less well served by research to date. This research project conducts an investigation into the science underlying the implementation of portfolio management strategies in practice focusing on enhanced indexing strategies. Enhanced indexing is a passive management approach which introduces an element of active management with the aim of achieving a level of active return through small adjustments to the portfolio weights. …

Can A Strictly Defined Security Configuration For Iot Devices Mitigate The Risk Of Exploitation By Botnet Malware?, David Kennefick

Dissertations

The internet that we know and use every day is the internet of people, a collection of knowledge and data that can be accessed anywhere is the world anytime from many devices. The internet of the future is the Internet of Things. The Internet of Things is a collection of automated technology that is designed to be run autonomously, but on devices designed for humans to use. In 2016 the Mirai malware has shown there are underlying vulnerabilities in devices connected to the internet of things. Mirai is specifically designed to recognise and exploit IoT devices and it has been …