Engineering Commons^™

Investigating The Impact Of Unsupervised Feature-Extraction From Multi-Wavelength Image Data For Photometric Classification Of Stars, Galaxies And Qsos, Annika Lindh

Conference papers

Accurate classification of astronomical objects currently relies on spectroscopic data. Acquiring this data is time-consuming and expensive compared to photometric data. Hence, improving the accuracy of photometric classification could lead to far better coverage and faster classification pipelines. This paper investigates the benefit of using unsupervised feature-extraction from multi-wavelength image data for photometric classification of stars, galaxies and QSOs. An unsupervised Deep Belief Network is used, giving the model a higher level of interpretability thanks to its generative nature and layer-wise training. A Random Forest classifier is used to measure the contribution of the novel features compared to a set …

Support Vector Machines And Artificial Neural Networks: Assessing The Validity Of Using Technical Features For Security Forecasting, James Dipadua

Dissertations

Stock forecasting is an enticing and well-studied problem in both finance and machine learning literature with linear-based models such as ARIMA and ARCH to non-linear Artificial Neural Networks (ANN) and Support Vector Machines (SVM). However, these forecasting techniques also use very different input features, some of which are seen by economists as irrational and theoretically unjustified. In this comparative study using ANNs and SVMs for 12 publicly traded companies, derivative price “technicals” are evaluated against macro- and microeconomic fundamentals to evaluate the efficacy of model performance. Despite the efficient market hypothesis positing the ill-suitability of technicals as model inputs, this …

Activist: A New Framework For Dataset Labelling, Jack O'Neill, Sarah Jane Delany, Brian Mac Namee

Conference papers

Acquiring labels for large datasets can be a costly and time-consuming process. This has motivated the development of the semi-supervised learning problem domain, which makes use of unlabelled data — in conjunction with a small amount of labelled data — to infer the correct labels of a partially labelled dataset. Active Learning is one of the most successful approaches to semi-supervised learning, and has been shown to reduce the cost and time taken to produce a fully labelled dataset. In this paper we present Activist; a free, online, state-of-the-art platform which leverages active learning techniques to improve the efficiency of …

Empirical Comparative Analysis Of 1-Of-K Coding And K-Prototypes In Categorical Clustering, Fei Wang, Hector Franco, John Pugh, Robert J. Ross

Conference papers

Clustering is a fundamental machine learning application, which partitions data into homogeneous groups. K-means and its variants are the most widely used class of clustering algorithms today. However, the original k-means algorithm can only be applied to numeric data. For categorical data, the data has to be converted into numeric data through 1-of-K coding which itself causes many problems. K-prototypes, another clustering algorithm that originates from the k-means algorithm, can handle categorical data by adopting a different notion of distance. In this paper, we systematically compare these two methods through an experimental analysis. Our analysis shows that K-prototypes is more …

Support Vector Machines And Artificial Neural Networks: Assessing The Validity Of Using Technical Features For Security Forecasting, James Di Padua

Dissertations

Stock forecasting is an enticing and well studied problem in both finance and machine learning literature with linear based models such as ARIMA and ARCH to nonlinear Artificial Neural Networks (ANN) and Support Vector Machines (SVM). However, these forecasting techniques also use very different input features, some of which are seen by economists as irrational and theoretically unjustified. In this comparative study using ANNs and SVMs for 12 publicly traded companies, derivative price “technicals” are evaluated against macro and microeconomic fundamentals to evaluate the efficacy of model performance. Despite the efficient market hypothesis positing the ill suitability of technicals as …

Using Spatialisation To Support Exploratory Search Behaviour, Clement Roux

Dissertations

Information-seekers traditionally interact with digital content through keyword-based search interfaces displaying results in list views. Well-defined lookup search tasks are performed brilliantly with these interfaces, enabling users to find relevant information and develop a relative understanding of the underlying information space. However, it is feasible to suggest that ill-defined and abstract search tasks could be better supported with a different interface that could allow the user to explore a library’s content and develop an appropriate mental model of the information space. One such approach is based on the use of visualisation, an approach to data analysis that aims to reduce …

A Regression Study Of Salary Determinants In Indian Job Markets For Entry Level Engineering Graduates, Rajveer Singh

Dissertations

The economic liberalisation of Indian markets in early 90s boosted the economic growth of the nation in various sectors over the next two decades. One such sector that has seen a massive growth in this time is Information Technology (IT). The IT industry has played a very crucial role in transforming India from a slow moving economy to one of the largest exporters of IT services. This growth created a huge demand in the labour markets for skilled labour, which in turn made engineering one of the top choices of study after high school over the years. In addition, the …

Investigating The Impact Of Unsupervised Feature-Extraction From Multi-Wavelength Image Data For Photometric Classification Of Stars, Galaxies And Qsos, Annika Lindh

Dissertations

This thesis reviews the current state of photometric classification in Astronomy and identifies two main gaps: a dependence on handcrafted rules, and a lack of interpretability in the more successful classifiers. To address this, Deep Learning and Computer Vision were used to create a more interpretable model, using unsupervised training to reduce human bias.

The main contribution is the investigation into the impact of using unsupervised feature-extraction from multi-wavelength image data for the classification task. The feature-extraction is achieved by implementing an unsupervised Deep Belief Network to extract lower-dimensionality features from the multi-wavelength image data captured by the Sloan Digital …

An Exploration Of The Impact Of Animal Agriculture On Human Sustainable Development Index And The Child Health Indicator, Hithesan Pandian

Dissertations

Several studies have shown that animal agriculture is one of the major contributors to climate change due to greenhouse gas emissions, deforestation, land degradation, freshwater shortages, general environmental pollution and world hunger. Apart from this, meat consumption is strongly associated with certain fatal health conditions such as cancer, cardiovascular disease and diabetes (Wu, 2014). Although meat production and export could be economically beneficial in the short run, it could lead to over-exploitation of natural resources and in turn the destruction of the environment in the long run. Hence, it is essential for the leaders of a nation to make smart …

Model-Free And Model-Based Active Learning For Regression, Jack O'Neill, Sarah Jane Delany, Brian Macnamee

Conference papers

Training machine learning models often requires large labelled datasets, which can be both expensive and time-consuming to obtain. Active learning aims to selectively choose which data is labelled in order to minimize the total number of labels required to train an effective model. This paper compares model-free and model-based approaches to active learning for regression, finding that model-free approaches, in addition to being less computationally intensive to implement, are more effective in improving the performance of linear regressions than model-based alternatives.

Activist: A New Framework For Dataset Labelling, Jack O'Neill, Sarah Jane Delany, Brian Macnamee

Conference papers

Acquiring labels for large datasets can be a costly and time-consuming process. This has motivated the development of the semi-supervised learning problem domain, which makes use of unlabelled data — in conjunction with a small amount of labelled data — to infer the correct labels of a partially labelled dataset. Active Learning is one of the most successful approaches to semi-supervised learning, and has been shown to reduce the cost and time taken to produce a fully labelled dataset. In this paper we present Activist; a free, online, state-of-the-art platform which leverages active learning techniques to improve the efficiency …

Lifelong Housing Design: User Feedback Evaluation Of Smart Objects And Accessible Houses For Healthy Ageing, Matteo Zallio, Niccolò Casiddu

Conference Papers

According to the latest research by the European Community and ISTAT (Italian National Institute of Statistics) surveys, Europe has the highest average age for its population. According to those data, in the near future, it could be necessary to move from a welfare model based on the centralization of care systems, to a system based on the distribution of certain healthcare facilities [1]. This means that the ageing population is ever increasing, thanks to better lifestyles, innovative medical care and wider access to different services. This work seeks to observe and analyse key implications of architectural and interior design features …

Bitrate Classification Of Twice-Encoded Audio Using Objective Quality Features, Colm Sloan, Damien Kelly, Naomi Harte, Anil C. Kokaram, Andrew Hines

Conference papers

When a user uploads audio files to a music stream- ing service, these files are subsequently re-encoded to lower bitrates to target different devices, e.g. low bitrate for mobile. To save time and bandwidth uploading files, some users encode their original files using a lossy codec. The metadata for these files cannot always be trusted as users might have encoded their files more than once. Determining the lowest bitrate of the files allows the streaming service to skip the process of encoding the files to bitrates higher than that of the uploaded files, saving on processing and storage space. This …

An Investigation Into The Effectiveness Of Hedonic Features In Regression Models For Domestic Rental Prices, Peter Prunty

Dissertations

Housing is a fundamental human right. Increasing rents and rising unemployment contribute to increased rates of homelessness. Traditionally housing prices are determined by supply and demand. This project will investigate the relationship between hedonic features and domestic rental prices in California and New York, using multivariate regression models. The literature outlines a number of approaches taken to model real estate pricing using hedonic regression.

Two models were created to analyse the difference between California and New York. Features were selected using correlation analysis. Some features were derived using logarithmic and dummy feature transformations. The models themselves were evaluated by assessing …

Investigation Into The Predictive Capability Of Macro-Economic Features In Modelling Credit Risk For Small Medium Enterprises, Kevin Mctiernan

Dissertations

This research project investigates the predictive capability of macro-economic features in modelling credit risk for small medium enterprises (SME/SMEs). There have been indications that there is strong correlation between economic growth and the size of the SME sector in an economy. However, since the financial crisis and consequent policies and regulations, SMEs have been hampered in attempts to access credit. It has also been noted that while there is a substantial amount of credit risk literature, there is little research on how macro-economic factors affect credit risk. Being able to improve credit scoring by even a small amount can have …

Using A Knowledge Management Approach To Support Effective Succession Planning In The Civil Service, Mary O’Donohue

Dissertations

The modern workforce is highly mobile. The challenge facing organisations is how to safeguard key expertise and knowledge in the face of staff mobility and turnover. The Irish Civil Service is still recovering from the impacts of significant loss of staff, and their knowledge and expertise, as a result of cutbacks over recent years. This project will establish the potential of using a Knowledge Management approach to support effective succession planning in the Civil Service. The literature review charts the evolution of Knowledge Management from when the phrase was first coined in 1986 through to what is considered to be …

An Exploration Of The Relationship Between The Partisan-Business Cycle And Economic Inequality Within Developed Economies, Richard O'Doherty

Dissertations

Recent contributions to the study of inequality have provided strong evidence towards the presence of an established trend, over several decades, of growing economic inequality (with a particular focus on distribution within their tails; i.e. top 10%, 1%) across countries with developed economies and indications of similar trends across developing economies. While the causality and influencing factors to these trends has widely been discussed, and has range from declining domestic growth rates as economies move towards high mass consumption states to globalisation, political decision making and policy application been referred to as both contributory or an instrument for dampening such …

Source Separation Approach To Video Quality Prediction In Computer Networks, Ruairí De Fréin

Articles

Time-varying loads introduce errors in the estimated model parameters of service-level predictors in Computer Networks. A load-adjusted modification of a traditional unadjusted service-level predictor is contributed, based on Source Separation (SS). It mitigates these errors and improves service-quality predictions for Video-on-Demand (VoD) by :6 to 2dB.

Measuring The Effectiveness Of Software-Based Training To Improve The Spatial Visualization Skills Of Students In Stem Disciplines In Higher Education Institutions, Peter Cole

Dissertations

This research investigates how software can be used to teach spatial skills leading to greater success in Science, Technology, Engineering, and Mathematical (STEM) fields. Existing research indicates that spatial skills can be taught and that good spatial skills are common to people who succeed in STEM fields. In this work, a software-only testing system with a direct targeted, training intervention module was implemented to measure and teach spatial skills using mental rotations, which are believed to be one of the most significant indicators of success in STEM fields. Spatial skills were tested using a standardized and validated test that measures …

Forecasting Bike Rental Demand Using New York Citi Bike Data, Wen Wang

Dissertations

The idea of this project is from a Kaggle competition “Bike Sharing Demand”① which provides dataset of Capital Bikeshare in Washington D.C. and asked to combine historical usage patterns with weather data in order to forecast bike rental demand. This dissertation will extend this work, working with a broader range of project not only just focusing on the phrase of model building but all phases of KDD (Knowledge Discovery in Databases). This dissertation focuses on Citi Bike which is one of the biggest bike share projects in the world, collects Citi Bike data, weather data and holiday data from three …

An Evaluation Of Gamification To Assess Students’ Learning On Their Understanding Of First Year Computer Science Programming Module, Daniel Gebremichael

Dissertations

This research examines the use of gamification to develop an assessment tool, to assess students’ learning of a first year computer science module. The students’ undertaking of the first semester Programming and Algorithms module in 2015 were assessed on their knowledge of the programming language Python. The incorporation of gamification when assessing students can have various potential benefits. The research aims to identify these benefits and issues. Assessments and games have almost opposite effects on opinions on people, as games are usually expected to have an entertainment value but this is not the case for assessments. The research examines if …

How Do Teachers And Students Perceive The Utility Of Blackboard As A Distance Learning Platform? (Case Study From Taibah University, Saudi Arabia), Suad Alaofi

Dissertations

This research explores the role of Knowledge Management within the education field with a specific focus on the use of Learning Management Systems in the Distance Learning (e-learning) process. The aim of this study is to thoroughly examine how teachers and students perceive the utility of the Blackboard system as a distance learning platform. To achieve this, the study conducted qualitative interviews and quantitative questionnaire surveys with the teachers and students of Taibah University, Saudi Arabia. Questions in both data collection tools were geared towards gaining insight about how these two groups of Blackboard users view its usefulness as a …

Simultaneous Rotation And Distance Measurement Using Multiband Circularly Polarized Radio Link, Adam Narbudowicz

Conference Papers

The paper proposes a method for simultaneous measurement of rotation and distance by using a multi-band circularly polarized radio links. The method is realized using a standard dual-band patch antenna, operating at 1.58 GHz and 2.27 GHz, with respective axial ratios of 1.7 dB and 1.0 dB. Both link distance and antenna rotation were varied from a reference position/orientation from -40 mm to +40 mm and from 0° to 360°, producing a total set of 125 samples. For a noiseless link the distance change and rotation were predicted with respective mean errors of 1.9 mm and 7.3°.

Predicting Intake Of Applications For First Registration In The Property Registration Authority, Orlaith Mernagh

Dissertations

The motivation for this dissertation is rooted in a real business need. The Property Registration Authority is the state organisation tasked with maintaining a register of land ownership on the island of Ireland. The PRA currently faces a series of challenges; a high level of staff retiring and the inherent loss of knowledge associated with this trend, a lack of recruitment in recent years and a large increase in lodgement of applications for first registration as a result of legislation. The organisation therefore requires a reliable system for predicting future intake. Prior to this project, there has also been a …

Identifying Market Indicators And Content Quality From A Financial Micro-Blog Platform, Siobhán Mcnamara

Dissertations

Investment platforms and discussion platforms have come to change the face of finance. The stock market is open to both professional and non-professional investors via online financial channels. Information too comes via a shared domain as both professionals and non-professionals log onto online communication platforms to share, search and discuss market trends. Due to their growing role in finance, understanding online communities has become the focus of much stock market research. Determining who is influential in a network, how information spreads and what translates to buy or sell decision is potentially very lucrative. In this research paper a dataset from …

Modeling Mental Workload Via Rule-Based Expert System: A Comparison With Nasa-Tlx & Workload Profile, Lucas Rizzo, Sarah Jane Delany, Pierpaolo Dondio, Luca Longo

Conference papers

In the last few decades several fields have made use of the construct of human mental workload (MWL) for system and task design as well as for assessing human performance. Despite this interest, MWL remains a nebulous concept with multiple definitions and measurement techniques. State-of-the-art models of MWL are usually ad-hoc, considering different pools of pieces of evidence aggregated with different inference strategies. In this paper the aim is to deploy a rule-based expert system as a more structured approach to model and infer MWL. This expert system is built upon a knowledge-base of an expert and transates into computable …

Monitoring Voip Speech Quality For Chopped And Clipped Speech, Andrew Hines, Jan Skoglund, Anil C. Kokaram, Naomi Harte

Articles

No abstract provided.

A New Educational Mobile Devices Platform For Social Inclusion In Tanzania, Fredrick Mtenzi

Articles

Abstract— It is evident that advances in technology has led to improvement in societal wellbeing. In this paper we demonstrate how mobile phones are used in providing reliable and quality education to students in disadvantaged areas of Tanzania. The main contribution is on leveraging on the success that Tanzania has had on using mobile banking to the un-banked population. These lessons are adapted to the education sector, where clever/smart integration of existing disruptive technologies such as mobile phones and social networks are be used to provide access to high quality educational contents. Further, the paper shows how educational content can …

Kicm: A Knowledge-Intensive Context Model, Fredrick Mtenzi, Denis Lupiana

Conference papers

A context model plays a significant role in developing context-aware architectures and consequently on realizing context-awareness, which is important in today's dynamic computing environments. These architectures monitor and analyse their environments to enable context-aware applications to effortlessly and appropriately respond to users' computing needs. These applications make the use of computing devices intuitive and less intrusive. A context model is an abstract and simplified representation of the real world, where the users and their computing devices interact. It is through a context model that knowledge about the real world can be represented in and reasoned by a context-aware architecture. This …