Stock market prediction has been a hot topic lately due to advances in computer technology and economics. One economic theory, called Efficient Market Hypothesis (EMH), states that all known information is already factored into the prices which makes it impossible to predict the stock market. Despite the EMH, many researchers have been successful in predicting the stock market using neural networks on historical data. This thesis investigates stock prediction using both linear regression and neural networks (NN), with a twist. The inputs to the proposed methods are a number of profit predictions calculated with stochastic methods such as generalized autoregressive conditional heteroskedasticity (GARCH) and autoregressive integrated moving average (ARIMA). By contrast the traditional approach was instead to use raw data as inputs. The proposed methods show superior result in yielding profit: at best 1.1% in the Swedish market and 4.6% in the American market. The neural network yielded more profit than the linear regression model, which is reasonable given its ability to find nonlinear patterns. The historical data was used with different window sizes. This gives a good understanding of the window size impact on the prediction performance.

The hypothesis that similar chemicals exert similar biological activities has been widely adopted in the field of drug discovery and development. Quantitative Structure-Activity Relationship (QSAR) models have been used ubiquitously in drug discovery to understand the function of chemicals in biological systems. A common QSAR modeling method calculates similarity scores between chemicals to assess their biological function. However, due to the fact that some chemicals can be similar and yet have different biological activities, or conversely can be structurally different yet have similar biological functions, various methods have instead been developed to quantify chemical similarity at the functional level. Quantmap is one such method, which utilizes biological databases to quantify the biological similarity between chemicals. Quantmap uses quantitative molecular network topology analysis to cluster chemical substances based on their bioactivities. This method by itself, unfortunately, cannot assign new chemicals (those which may not yet have biological data) to the derived clusters. Owing to the fact that there is a lack of biological data for many chemicals, deep learning models were explored in this project with respect to their ability to correctly assign unknown chemicals to Quantmap clusters. The deep learning methods explored included both convolutional and recurrent neural networks. Transfer learning/pretraining based approaches and data augmentation methods were also investigated. The best performing model, among those considered, was the Seq2seq model (a recurrent neural network containing two joint networks, a perceiver and an interpreter network) without pretraining, but including data augmentation.

Doctor of Philosophy / Department of Civil Engineering / Yacoub M. Najjar / This study explored the potential use of ANNs for profiling and characterization of various environmental sites. A static ANN with back-propagation algorithm was used to model the environmental containment at a hypothetical data-rich contaminated site. The performance of the ANN profiling model was then compared with eight known profiling methods. The comparison showed that the ANN-based models proved to yield the lowest error values in the 2-D and 3-D comparison cases. The ANN-based profiling models also produced the best contaminant distribution contour maps when compared to the actual maps. Along with the fact that ANN is the only profiling methodology that allows for efficient 3-D profiling, this study clearly demonstrates that ANN-based methodology, when properly used, has the potential to provide the most accurate predictions and site profiling contour maps for a contaminated site. ANN with a back-propagation learning algorithm was utilized in the site characterization of contaminants at the Kansas City landfill. The use of ANN profiling models made it possible to obtain reliable predictions about the location and concentration of lead and copper contamination at the associated Kansas City landfill site. The resulting profiles can be used to determine additional sampling locations, if needed, for both groundwater and soil in any contaminated zones. Back-propagation networks were also used to characterize the MMR Demo 1 site. The purpose of the developed ANN models was to predict the concentrations of perchlorate at the MMR from appropriate input parameters. To determine the most-appropriate input parameters for this model, three different cases were investigated using nine potential input parameters. The ANN modeling used in this case demonstrates the neural network’s ability to accurately predict perchlorate contamination using multiple variables. When comparing the trends observed using the ANN-generated data and the actual trends identified in the MMR 2006 System Performance Monitoring Report, both agree that perchlorate levels are decreasing due to the use of the Extraction, Treatment, and Recharge (ETR) systems. This research demonstrates the advantages of ANN site characterization modeling in contrast with traditional modeling schemes. Accordingly, characterization task-related uncertainties of site contaminations were curtailed by the use of ANN-based models.

Forecasting is used for decision making and unreliable predictions can instill a false sense of condence. Traditional time series modelling is astatistical art form rather than a science and errors can occur due to lim-itations of human judgment. In minimizing the risk of falsely specifyinga process the practitioner can make use of machine learning models. Inan eort to nd out if there's a benet in using models that require lesshuman judgment, the machine learning models Random Forest and Neural Network have been used to model a VAR(1) time series. In addition,the classical time series models AR(1), AR(2), VAR(1) and VAR(2) havebeen used as comparative foundation. The Random Forest and NeuralNetwork are trained and ultimately the models are used to make pre-dictions evaluated by RMSE. All models yield scattered forecast resultsexcept for the Random Forest that steadily yields comparatively precisepredictions. The study shows that there is denitive benet in using Random Forests to eliminate the risk of falsely specifying a process and do infact provide better results than a correctly specied model.

Published Article / An artificial neural network (ANN) that is able to distinguish between Cheddar cheese produced with milk from mixed and single breed sources was designed. Samples of each batch (4 pure Ayrshire/4 mixed with no Ayrshire milk) were ripened for 92 days and analysed every 14 days. A novel ANN was designed and applied which, based only on Lactococcus lactis counts, provided an acceptable classification of the cheeses. The ANN consisted of a multi-layered network with supervised training arranged in an ordered hierarchy of layers, in which connections were allowed only between nodes in immediately adjacent layers.

There are two main defined classes of uncertainty namely: fuzziness and ambiguity, where ambiguity is ‘one-to-many’ relationship between syntax and semantic of a proposition. This definition seems that it ignores ‘many-to-many’ relationship ambiguity type of uncertainty. In this thesis, we shall use complex-uncertainty to term many-to-many relationship ambiguity type of uncertainty. This research proposes a new approach for handling the complex ambiguity type of uncertainty that may exist in data, for software engineering of predictive Data Mining (DM) classification models. The proposed approach is based on Relative-Fuzzy Logic (RFL), a novel type of fuzzy logic. RFL defines a new formulation of the problem of ambiguity type of uncertainty in terms of States Of Proposition (SOP). RFL describes its membership (semantic) value by using the new definition of Domain of Proposition (DOP), which is based on the relativity principle as defined by possible-worlds logic. To achieve the goal of proposing RFL, a question is needed to be answered, which is: how these two approaches; i.e. fuzzy logic and possible-world, can be mixed to produce a new membership value set (and later logic) that able to handle fuzziness and multiple viewpoints at the same time? Achieving such goal comes via providing possible world logic the ability to quantifying multiple viewpoints and also model fuzziness in each of these multiple viewpoints and expressing that in a new set of membership value. Furthermore, a new architecture of Hierarchical Neural Network (HNN) called ML/RFL-Based Net has been developed in this research, along with a new learning algorithm and new recalling algorithm. The architecture, learning algorithm and recalling algorithm of ML/RFL-Based Net follow the principles of RFL. This new type of HNN is considered to be a RFL computation machine. The ability of the Relative Fuzzy-based DM prediction model to tackle the problem of complex ambiguity type of uncertainty has been tested. Special-purpose Integrated Development Environment (IDE) software, which generates a DM prediction model for speech recognition, has been developed in this research too, which is called RFL4ASR. This special purpose IDE is an extension of the definition of the traditional IDE. Using multiple sets of TIMIT speech data, the prediction model of type ML/RFL-Based Net has classification accuracy of 69.2308%. This accuracy is higher than the best achievements of WEKA data mining machines given the same speech data.

Borehole imaging is one of the fastest and most precise methods for collecting subsurface data that provides high resolution information on layering, texture and dips, permitting a core-like description of the subsurface. Although the range of information recoverable from this technology is widely acknowledged, image logs are still used in a strictly qualitative manner. Interpreting image logs manually is cumbersome, time consuming and is subjective based on the experience of the interpreter. This thesis outlines new methods that automate image log interpretation and extract subsurface lithofacies information in a quantitative manner. We developed two methodologies based on advanced image analysis techniques successfully employed in remote sensing and medical imaging. The first one is a pixelbased pattern recognition technique applying textural analysis to quantify image textural properties. These properties together with standard logs and core-derived lithofacies information are used to train a back propagation Neural Network. In principle the trained and tested Neural Network is applicable for automated borehole image interpretation from similar geological settings. However, this pixel-based approach fails to make use explicitly of the spatial characteristics of a high resolution image. TAT second methodology is introduced which groups identical neighbouring pixels into objects. The resultant spectrally and spatially consistent objects are then related to geologically meaningful groups such as lithofacies by employing fuzzy classifiers. This method showed better results and is applied to outcrop photos, core photos and image logs, including a ‘difficult’ data set from a deviated well. The latter image log did not distinguish some of the conductive and resistive regions, as observed from standard logs and core photos. This is overcome by marking bed boundaries using standard logs. Bed orientations were estimated using an automated sinusoid fitting algorithm within a formal uncertainty framework in order to distinguish dipping beds and horizontal stratification. Integration of these derived logs in the methodology yields a complete automated lithofacies identification, even from the difficult dataset. The results were validated through the interpretation of cored intervals by a geologist. This is a supervised classification method which incorporates the expertise of one or several geologists, and hence includes human logic, reasoning, and current knowledge of the field heterogeneity. By including multiple geologists in the training, the results become less dependent on each individual’s subjectivity and prior experience. The method is also easily adaptable to other geological settings. In addition, it is applicable to several kinds of borehole images, for example wireline electrical borehole wall images, core photographs, and logging-while-drilling (LWD) images. Thus, the theme of this dissertation is the development of methodologies which makes image log interpretation simpler, faster, less subjective, and efficient such that it can be applied to large quantities of data.

Dementia is increasing recognized as a major problem to public health worldwide. Prevention and treatment strategies are in critical need. Nowadays, research for dementia usually featured as complex longitudinal studies, which provide extensive information and also propose challenge to statistical methodology. The purpose of this dissertation research was to apply statistical methodology in the field of dementia to strengthen the understanding of dementia from three perspectives: 1) Application of statistical methodology to investigate the association between potential risk factors and incident dementia. 2) Application of statistical methodology to analyze changes over time, or trajectory, in cognitive tests and symptoms. 3) Application of statistical learning methods to predict development of dementia in the future. Prevention of Alzheimer’s disease with Vitamin E and Selenium (PREADViSE) (7547 subjects included) and Alzheimer’s disease Neuroimaging Initiative (ADNI) (591 participants included) were used in this dissertation. The first study, “Self-reported sleep apnea and dementia risk: Findings from the PREADViSE Alzheimer’s disease prevention trial ”, shows that self-reported baseline history of sleep apnea was borderline significantly associated with risk of dementia after adjustment for confounding. Stratified analysis by APOE ε4 carrier status showed that baseline history of sleep apnea was associated with significantly increased risk of dementia in APOE ε4 non-carriers. The second study, “comparison of trajectories of episodic memory for over 10 years between baseline normal and MCI ADNI subjects,” shows that estimated 30% normal subjects at baseline assigned to group 3 and 6 stay stable for over 9 years, and normal subjects at baseline assigned to Group 1 (18.18%) and Group 5 (16.67%) were more likely to develop into dementia. In contrast to groups identified for normal subjects, all trajectory groups for MCI subjects at baseline showed the tendency to decline. The third study, “comparison between neural network and logistic regression in PREADViSE trial,” demonstrates that neural network has slightly better predictive performance than logistic regression, and also it can reveal complex relationships among covariates. In third study, the effect of years of education on response variable depends on years of age, status of APOE ɛ4 allele and memory change.

Aeroderivative gas turbines are used all over the world for different applications as Combined Heat and Power (CHP), Oil and Gas, ship propulsion and others. They combine flexibility with high efficiencies, low weight and small footprint, making them attractive where power density is paramount as off shore Oil and Gas or ship propulsion. In Western Europe they are widely used in CHP small and medium applications thanks to their maintainability and efficiency. Reliability, Availability and Performance are key parameters when considering plant operation and maintenance. The accurate diagnose of Performance is fundamental for the plant economics and maintenance planning. There has been a lot of work around units like the LM2500® , a gas generator with an aerodynamically coupled gas turbine, but nothing has been found by the author for the LM6000® . Water wash, both on line or off line, is an important maintenance practice impacting Reliability, Availability and Performance. This Thesis aims to select and apply a suitable diagnostic technique to help establishing the schedule for off line water wash on a specific model of this engine type. After a revision of Diagnostic Methods Artificial Neural Network (ANN) has been chosen as diagnostic tool. There was no WebEngine model available of the unit under study so the first step of setting the tool has been creating it. The last step has been testing of ANN as a suitable diagnostic tool. Several have been configured, trained and tested and one has been chosen based on its slightly better response. Finally, conclusions are discussed and recommendations for further work laid out.

This thesis presents a novel method of locating imbalance faults in rotating machinery through the study of bearing nonlinearities. Localisation in this work is presented as determining which discs/segments of a complex machine are affected with an imbalance fault. The novel method enables accurate localisation to be achieved using a single accelerometer, and is valid for both sub and super-critical machine operations in the presence of misalignment and rub faults. The development of the novel system for imbalance localisation has been driven by the desire for improved maintenance procedures, along with the increased requirement for Integrated Vehicle Health Management (IVHM) systems for rotating machinery in industry. Imbalance faults are of particular interest to aircraft engine manufacturers such as Rolls Royce plc, where such faults still result in undesired downtime of machinery. Existing methods of imbalance localisation have yet to see widespread implementation in IVHM and Engine Health Monitoring (EHM) systems, providing the motivation for undertaking this project. The imbalance localisation system described has been developed primarily for a lab-based Machine Fault Simulator (MFS), with validation and verification performed on two additional test rigs. Physics based simulations have been used in order to develop and validate the system. An Artificial Neural Network (ANN) has been applied for the purposes of reasoning, using nonlinear features in the frequency domain originating from bearing nonlinearities. The system has been widely tested in a range of situations, including in the presence of misalignment and rub faults and on a full scale aircraft engine model. The novel system for imbalance localisation has been used as the basis for a methodology aimed at localising common faults in future IVHM systems, with the aim of communicating the results and findings of this research for the benefit of future research. The works contained herein therefore contribute to scientific knowledge in the field of IVHM for rotating machinery.