Saudi Cultural Missions Theses & Dissertations
Permanent URI for this communityhttps://drepo.sdl.edu.sa/handle/20.500.14154/10
Browse
8 results
Search Results
Item Restricted Deep Learning Approaches for Multivariate Time Series: Advances in Feature Selection, Classification, and Forecasting(New Mexico State University, 2024) Alshammari, Khaznah Raghyan; Tran, Son; Hamdi, Shah MuhammadIn this work, we present the latest developments and advancements in the machine learning-based prediction and feature selection of multivariate time series (MVTS) data. MVTS data, which involves multiple interrelated time series, presents significant challenges due to its high dimensionality, complex temporal dependencies, and inter-variable relationships. These challenges are critical in domains such as space weather prediction, environmental monitoring, healthcare, sensor networks, and finance. Our research addresses these challenges by developing and implementing advanced machine-learning algorithms specifically designed for MVTS data. We introduce innovative methodologies that focus on three key areas: feature selection, classification, and forecasting. Our contributions include the development of deep learning models, such as Long Short-Term Memory (LSTM) networks and Transformer-based architectures, which are optimized to capture and model complex temporal and inter-parameter dependencies in MVTS data. Additionally, we propose a novel feature selection framework that gradually identifies the most relevant variables, enhancing model interpretability and predictive accuracy. Through extensive experimentation and validation, we demonstrate the superior performance of our approaches compared to existing methods. The results highlight the practical applicability of our solutions, providing valuable tools and insights for researchers and practitioners working with high-dimensional time series data. This work advances the state of the art in MVTS analysis, offering robust methodologies that address both theoretical and practical challenges in this field.14 0Item Restricted Reducing Type 1 Childhood Diabetes in Saudi Arabia by Identifying and Modelling Its Key Performance Indicators(Royal Melbourne Institute of Technology, 2024-06) Alazwari, Ahood; Johnstone, Alice; Abdollahain, Mali; Tafakori, LalehThe increasing incidence of type 1 diabetes (T1D) in children is a growing global health concern. Reducing the incidence of diabetes generally is one of the goals in the World Health Organisation’s (WHO) 2030 Agenda for Sustainable Development Goals. With an incidence rate of 31.4 cases per 100,000 children and an estimated 3,800 new cases per year, Saudi Arabia is ranked 8th in the world for number of T1D cases and 5th for incidence rate. Despite the remarkable increase in the incidence of childhood T1D in Saudi Arabia, there is a lack of meticulously carried out research on T1D in children when compared with developed countries. In addition, it is crucial to recognise the critical gaps in current understanding of diabetes in children, adolescents, and young adults, with recent research indicates significant global and sub-national variations in disease incidence. Better knowledge of the development of T1D in children and its associated factors would aid medical practitioners in developing intervention plans to prevent complications and address the incidence of T1D. This study employed statistical, machine learning and classification approaches to analyse and model different aspects of childhood T1D using local case and control data. In this study, secondary data from 1,142 individual medical records (359-377 cases and 765 controls) collected from three cities located in different regions of Saudi Arabia have been used in the analysis to represent the country’s diverse population. Case and control data matched by birth year, gender and location were used to control confounders and create a more robust and clinically relevant model. It is well documented that genetic and environmental factors contribute to childhood T1D so a wide range of potential key performance indicators (KPIs) from the literature were included in this study. The collected data included information on socioeconomic status, potential genetic and environmental factors, and demographic data such as city of residence, gender and birth year. Several techniques, such as cross-validation, hyperparameter tuning and bootstrapping, were used in this study to develop models. Common statistical metrics (coefficient of determination, R-squared, root mean squared error, mean absolute error) were used to evaluate performance for the regression models while for the classification models accuracy, sensitivity, precision, F score and area under the curve were utilised as performance measures. Multiple linear regression (MLR), artificial neural network (ANN) and random forest (RF) models were developed to predict the age at onset of T1D for all children 0-14 years old, as well as for the most common age group for onset, the 5-9 year olds. To improve the performance of the MLR models, interactions between variables were considered. Additionally, risk factors associated with the age at onset of T1D were identified. The results showed that MLR and RF outperformed ANN. The logarithm of age at onset was the most suitable dependent variable. RF outperformed the others for the 5-9 years age group. Birth weight, current weight and current height influenced the age at onset in both age groups. However, preterm birth was significant only in the 0-14 years cohort, while consanguineous parents and gender were significant in the 5-9 age group. Logistic regression (LR), random forest (RF), support vector machine (SVM), Naive Bayes (NB) and artificial neural network (ANN) models were utilised with case and control data to model the development of childhood T1D and to identify its key performance indicators. Full and reduced models were developed to determine the best model. The reduced models were built using the significant factors identified by the individual full model. The study found that full LR had the highest accuracy. Full RF and SVM with a linear kernel also performed well. Significant risk factors identified as being associated with developing childhood T1D include early exposure to cow’s milk, high birth weight, positive family history of T1D and maternal age over 25 years. Poisson regression (PR), RF, SVM and K-nearest neighbor (KNN) were then used to model the incidence of childhood T1D, taking in the identified significant risk factors. The interactions between variables were also considered to enhance the performance of the models. Both full and reduced models were created and compared to find the best models with the minimum number of variables. The full Poisson regression and machine learning models outperformed all other models, but reduced models with a combination of only two out of three independent variables (early exposure to cow’s milk, high birth weight and maternal age over 25 years) also performed relatively well. This study also deployed optimisation procedures with the reduced incidence models to develop upper and lower yearly profile limits for childhood T1D incidence to achieve the United Nations (UN) and Saudi recommended levels of 264 and 339 cases by 2030. The profile limits for childhood T1D then allowed us to model optimal yearly values for the number of children weighing more than 3.5kg at birth, the number of deliveries by older mothers and the number of children introduced early to cow’s milk. The results presented in this thesis will guide healthcare providers to collect data to monitor the most influential KPIs. This would enable the initiation of suitable intervention strategies to reduce the disease burden and potentially slow the incidence rate of childhood T1D in Saudi Arabia. The research outcomes lead to recommendations to establish early intervention strategies, such as educational campaigns and healthy lifestyle programs for mothers along with child health mentoring during and after pregnancy to reduce the incidence of childhood T1D. This thesis has contributed to new knowledge on childhood T1D in Saudi Arabia by: * developing a predictive model for age at onset of childhood T1D using statistical and machine learning models. * predicting the development of T1D in children using matched case-control data and identifying its KPIs using statistical and machine learning approaches. * modeling the incidence of childhood T1D using its associated significant KPIs. * developing three optimal profile limits for monitoring the yearly incidence of childhood T1D and its associated significant KPIs. * providing a list of recommendations to establish early intervention strategies to reduce the incidence of childhood T1D.14 0Item Restricted Synchronization, Learning and Classification for a System of Kuramoto Models(University of Exeter, 2024) Alanazi, Faizah; Townley, Stuart; Mueller, MarkusKuramoto systems or Kuramoto networks model the behaviour of large sets of coupled oscillators. Arising initially in the context of systems of chemical and biological oscillators, they now find applications in various areas of science, engineering and medicine, including neuroscience. A key property of Kuramoto networks is their synchronization behaviour: for a network with N oscillators, it is possible that all N oscillators synchronize, that several clusters of synchronized oscillators emerge, or that there is no synchronization of the oscillators. The behaviour is a function of the network parameters, namely coupling strengths and natural frequencies for the oscillators, as well as their initial conditions. In this thesis, we consider control systems theory approaches for Kuramoto networks that focus on adaptive learning of system parameters and phase tracking, observation-based classification of synchrony, and a combination of both. We first consider synchronization and learning approaches for pairs of Kuramoto networks. One network plays the role of a training network, the other is a learning network. We consider synchronization and system parameter learning based on phase information. Our main result is an adaptive learning strategy that tunes the system parameters of the learning oscillator – the Kuramoto coupling strengths and the natural frequencies – to achieve phase tracking, i.e. synchronization, between the training and learning phases. Tracking is proved using a Lyapunov stability approach. The adaptive strategy also guarantees partial convergence of the learning weights and frequencies to those of the training oscillator. Partial convergence is characterized by the linear dependence of the phase differences of the states of the training oscillator. The results are illustrated by a Kuramoto network with N = 4 oscillators. Secondly, and generalizing the synchronization and learning result, we consider networks where only output information is available and not all phases of the network i may be measured independently. A crucial aspect of this approach is the concept of observability and observer design for dynamical systems, i.e. how to make use of output information to recreate phase information. This is an unsolved problem for Kuramoto networks where a training system is not necessarily in an equilibrium state. To overcome this problem we develop a machine learning-based approach using so-called “fingerprints” of the networks output signals, i.e. spectrogram images that represent the possible synchronization behaviours. We use a simple artificial neural network architecture to develop a pattern recognition tools that classifies the “fingerprints” and thus the types synchrony as observed by outputs of Kuramoto networks of a fixed size. The approach is illustrated by simulation and classification results for Kuramoto networks with N = 4 and N = 7 oscillators. Using the classifier approach we then develop a switched systems adaptive control framework to determine the type of Kuramoto network responsible or able to create a given “fingerprint” that matches the “fingerprint” of the training system. Our second main result is an adaptive algorithm that can learn the behaviour of a Kuramoto network, from a set or family of possible networks, to match the output-based “fingerprint” of the training system. The results are illustrated for networks of N = 4 and N = 7 oscillators with a variety of synchrony outputs, respectively26 0Item Restricted Predicting Customer Attrition in B2B SaaS Using Machine Learning Classification(Saudi Digital Library, 2023-09-15) Alalawi, Zainab; Fiaschetti, MaurizioCustomer retention and customer loss are crucial metrics in subscription-based industries like SaaS companies. Customer discharge is a significant concern for this type of business, as clients have the flexibility to terminate the service at any time. This can lead to adverse effects on the company’s revenue stream. If SaaS businesses can accurately predict the number of customers who will cancel their subscriptions and those who will continue using their services within a specific timeframe, they can more effectively forecast their revenue, cash flow, and any future growth plan accordingly. Predicting subscription renewals and cancelations remains a challenging problem for any SaaS company. However, with the ongoing advancement in machine learning and artificial intelligence, the potential for accurately forecasting this issue has significantly improved. The study examines customer attrition and customer retention prediction in a quantitative method by utilizing several different machine learning algorithms with Python, namely Logistics regression, Naïve Baye, and random forest algorithms. Data was collected from the case company’s database and manipulated to fit the algorithms. The dataset includes the customers' business data such as spend, customer platform usage data, customer service history data, and the date of the next payment. To identify the best hyperparameters for each machine- learning algorithm, A tuning technique, in particular Grid Search, was employed. Subsequently, the algorithm models were trained and assessed using optimized hyperparameters on the fitted data. Once the models were trained, they were applied to test data to obtain the analysis results. The model’s performance was measured on the quantitative model performance metrics. including F1-Score, Area under Curve (AUC), and Accuracy.37 0Item Restricted DEEP LEARNING APPROACH TO LARGE-SCALE SYSTEMS(2023) Altamimi, Abdulelah; Lagoa, ConstantinoThe significance of large-scale systems has increased recently due to the growth in data and the number of users. The computational cost of analyzing these high-dimensional systems due to the curse of dimensionality raises the urge for developing efficient approaches. Deep learning methods have the capability and scalability to process high-volume data with significantly lower computational complexity. In this work, deep learning algorithms are utilized to solve large-scale systems in different applications. We design and solve high-dimensional systems using tractable algorithms. In particular, the deep reinforcement learning method and deep neural network are employed in our work in maximizing problems and classification problems, respectively. Comparisons with conventional algorithms are performed for validation purposes. Moreover, this work proposes an approach to exploiting the knowledge of the physical structure of plants inspired by deep learning algorithms. An application in the forest management field considered in this work is a large-scale forest model for wildfire mitigation. A high-dimensional forest model is designed in the Markov decision process framework. The model includes the probability of wildfire occurrence in a large number of stands. The probability of wildfire in each stand is a function of wind direction, flammability, and the stand's timber volume. Wildfire reduction is achieved by maximizing the timber volume in the forest through management actions. A deep reinforcement learning approach, i.e., the actor-critic algorithm, is used to solve the Markov decision process and propose management policies. Furthermore, the performances of conventional Markov decision process solutions, i.e., the value iteration algorithm and the genetic algorithm, are compared to the proposed approach. It outperforms these algorithms in terms of the value of the timber volume and the computational cost. Another interesting application considered in this thesis is fast stochastic predictive control. In the proposed approach, the computational complexity of solving stochastic predictive control is significantly reduced using deep learning. In particular, the number of constraints in the sampled method is reduced to the minimal set required to solve the optimization problem. Determining these constraints,i.e., the policies, is considered a classification problem to be solved using a neural network. The small number of constraints and the solvable quadratic optimization problem introduced by the sampled method result in a fast stochastic model predictive control. In this thesis, we also propose an approach to exploiting the prior knowledge of the physically interconnected systems in the parameter estimation domain. Unlike the physics-informed neural network, the proposed approach can estimate the parameters for every system in the interconnection. It has a general form that can be applied to any system as well as an objective function. We also combine the case of prior knowledge of system function with the case of the unavailability of this information. The Fourier series approximation method is used when knowledge of system functions is not available. The first-order gradient descent algorithm is considered to minimize the estimation error in the objective function. For that, we provide a systematic way to compute the gradients of the objective function. Using several versions of the gradient descent algorithm, the proposed solution shows promising results in the estimation of the system parameters.33 0Item Restricted Deep Discourse Analysis for Early Prediction of Multi-Type Dementia(Saudi Digital Library, 2023-06-12) Alkenani, Ahmed Hassan A; Li, YuefengAgeing populations are a worldwide phenomenon. Although it is not an inevitable consequence of biological ageing, dementia is strongly associated with increasing age, and is therefore anticipated to pose enormous future challenges to public health systems and aged care providers. While dementia affects its patients first and foremost, it also has negative associations with caregivers’ mental and physical health. Dementia is characterized by irreversible gradual impairment of nerve cells that control cognitive, behavioural, and language processes, causing speech and language deterioration, even in preclinical stages. Early prediction can significantly alleviate dementia symptoms and could even curtail the cognitive decline in some cases. However, the diagnostic procedure is currently challenging as it is usually initiated with clinical-based traditional screening tests. Typically, such tests are manually interpreted and therefore may entail further tests and physical examinations thus considered timely, expensive, and invasive. Therefore, many researchers have adopted speech and language analysis to facilitate and automate its initial prescreening. Although recent studies have proposed promising methods and models, there is still room for improvement, without which automated pre-screening remains impracticable. There is currently limited empirical literature on the modelling of the discourse ability of people with prodromal dementia stages and types, which is defined as spoken and written conversations and communications. Specifically, few researchers have investigated the nature of lexical and syntactic structures in spontaneous discourse generated by patients with dementia under different conditions for automated diagnostic modelling. In addition, most previous work has focused on modelling and improving the diagnosis of Alzheimer’s disease (AD), as the most common dementia pathology, and neglect other types of dementia. Further, current proposed models suffer from poor performance, a lack of generalizability, and low interpretability. Therefore, this research thesis explores lexical and syntactic presentations in written and spoken narratives of people with different dementia syndromes to develop high-performing diagnostic models using fusions of different lexical and syntactic (i.e., lexicosyntactic) features as well as language models. In this thesis, multiple novel diagnostic frameworks are proposed and developed based on the “wisdom of crowds” theory, in which different mathematical and statistical methods are investigated and properly integrated to establish ensemble approaches for an optimized overall performance and better inferences of the diagnostic models. Firstly, syntactic- and lexical-level components are explored and extracted from the only two disparate data sources available for this study: spoken and written narratives retrieved from the well-known DementiaBank dataset, and a blog-based corpus collected as a part of this research, respectively. Due to their dispersity, each data source was independently analysed and processed for exploratory data analysis and feature extraction. One of the most common problems in this context is how to ensure a proper feature space is generated for machine learning modelling. We solve this problem by proposing multiple innovative ensemble-based feature selection pipelines to reveal optimal lexicosyntactics. Secondly, we explore language vocabulary spaces (i.e., n-grams) given their proven ability to enhance the modelling performance, with an overall aim of establishing two-level feature fusions that combine optimal lexicosyntactics and vocabulary spaces. These fusions are then used with single and ensemble learning algorithms for individual diagnostic modelling of the dementia syndromes in question, including AD, Mild Cognitive Impairment (MCI), Possible AD (PoAD), Frontotemporal Dementia (FTD), Lewy Body Dementia (LBD), and Mixed Dementia (PwD). A comprehensive empirical study and series of experiments were conducted for each of the proposed approaches using these two real-world datasets to verify our frameworks. Evaluation was carried out using multiple classification metrics, returning results that not only show the effectiveness of the proposed frameworks but also outperform current “state-of-the-art” baselines. In summary, this research provides a substantial contribution to the underlying task of effective dementia classification needed for the development of automated initial pre-screenings of multiple dementia syndromes through language analysis. The lexicosyntactics presented and discussed across dementia syndromes may highly contribute to our understanding of language processing in these pathologies. Given the current scarcity of related datasets, it is also hoped that the collected writing-based blog corpus will facilitate future analytical and diagnostic studies. Furthermore, since this study deals with associated problems that have been commonly faced in this research area and that are frequently discussed in the academic literature, its outcomes could potentially assist in the development of better classification models, not only for dementia but also for other linguistic pathologies.18 0Item Restricted Optimal 0−1 Loss Classification In Linear, Overlapping And Interpolation Settings(University of Birmingham, 2022-09-07) Alanazi, Reem; Max, LittleClassification problems represent a major subject in machine learning, and addressing them requires solving underlying optimization problems. Optimis- ing the 0–1 loss function is the natural pathway to address the classification problem. Because of the properties of 0–1 loss, which are that it is non-convex and non-differentiable, 0–1 loss is mathematically intractable and classified as non-deterministic polynomial-time hard (NP-hard). Consequently, the 0– 1 loss function has been replaced by surrogate loss functions that can be optimized more efficiently, where their optimal solution is guaranteed with respect to these surrogate losses. At the same time, these functions may not provide the same optimal solution as the 0–1 loss function. Indeed, the loss function used during the empirical risk minimization (ERM) is not the same loss function used in the evaluation; the mismatch of the loss functions leads to an approximate solution that may not be an ideal solution for 0–1 loss. Thus, an additional source of error is produced because of this replacement.16 0Item Restricted Using Deep Learning Techniques for an Early Detection of Oral Epithelial Dysplasia(2023) Aljuaid, Abeer; Anwar, MohdOral cancer is ranked as the sixth most common type of cancer worldwide, with 90% of cases being oral squamous cell carcinoma (OSCC). OSCC has a high mortality rate, and early diagnosis can increase the survival rate. About 80% of OSCC is developed from Oral Epithelial Dysplasia (OED); thus, OED detection is critical to diagnose OSCC at the early stage. Traditionally, the OED is defined by sixteen criteria, including architectural and cytological features, under the microscope by expert oral pathologists. This manual detection is a time-consuming and tedious task, and thus, there is a need for precise automated diagnostic and classification techniques. However, disengaging a Computer Aided Diagnosis (CAD) for OED is challenging because each OED’s criteria require a particular medical image processing task for detection. Therefore, we proposed a novel multi-task learning network to combine semantic segmentation and classification to detect OED using architectural and cytological characteristics. Our proposal is the first study that jointly trained semantic segmentation and classification on a single network for automated OED detection. We developed four new frameworks called VGG16-UNet, InceptionV3-UNet, DyspVGG16, and Dysp-InceptionV3. The VGG16-UNet and InceptionV3-UNet were designed based on classic U-Net with the ImageNet pre-trained VGG16 and InceptionV3 encoder and a traditional classifier model. We built Dysp-VGG16 and Dysp-InceptionV3 using our novel modified U-Net and novel classifier network. Our modified U-Net involved dilated convolution, channel attention, spatial attention, and residual blocks for performance enhancement. The proposed models’ effectiveness and robustness were verified by running three experiments and utilizing quantitative metrics and visualization results for comparison. Consequently, our novel modified U-Net and classifier network show superior performance on classification and segmentation tasks. Our novel classifier enhanced the quantitative metrics and reduced the traditional classifier’s false positives and negative rates. Modified U-Net improved the semantic segmentation performance by 5% of the Jaccard index and provided accurate predicted masks.28 0