Applied Modeling and Quantitative Methods

The Multidimensional Inventory of Personal Intelligence (MIPI) was designed to measure three related dimensions of the personal intelligence (PI) construct: emotional intelligence (EI), social intelligence (SI), and motivational intelligence (MI). The MIPI has psychometric properties and a theoretical structure that improves on the shortcomings of existing trait EI measures. The aim of the first study was to create and validate a shortened form (MIPI- Short) that maintains the same factorial structure of the original MIPI. The purpose of the second study was to validate the new scale with measures of conceptually similar constructs (e.g., emotional intelligence, Alexithymia) with various measurement methodologies (self-report, observer-report, and performance-based). Results from Study 1 found that the MIPI-Short had good factorial structure in two independent samples, as well as adequate internal reliability, and good incremental validity. The results of Study 2 demonstrated that the MIPI-Short had good construct validity as it generally related as expected with measures of EI and Alexithymia. The findings of both studies provide evidence for the validity of the MIPI-Short as a brief measure of Personal Intelligence. Directions for further research are emphasized, as the validation process is on-going for any assessment tool.

Author Keywords: Emotional Intelligence, Personal Intelligence, Socio-Emotional Competencies

Objective: The Coping Inventory for Stressful Situations (CISS) is a widely used measure of trait coping that was developed to assess three basic coping styles: task-oriented, emotion-oriented, and avoidance-oriented coping. This thesis examined the psychometric properties of a short form for the CISS (CISS-SF). Method: Data from a large longitudinal sample of adults were used to conduct analyses testing the measure's factor structure, internal and test-retest reliabilities, and construct validity with respect to mental health outcomes. Results: The 3-factor model provided acceptable fit to the sample data. Internal reliabilities for the scales were acceptable across multiple administrations (by gender and age), while 1 and 2-year test-retest correlations were also consistent with what would be expected for stable coping style constructs. Relationships were found to be consistent with previous research on coping. Conclusion: Overall, the results suggest that the CISS-SF is a valid and reliable brief multi-dimensional measure of coping styles.

Author Keywords: basic personality, coping, coping styles, mental health, psychometrics

This study focuses on the world dynamics of a cholera model that includes delayed bacterial shedding and water disinfection. From the method of the next generation matrix, a basic reproduction number is found that sets a threshold of disease persistence. It is shown that the disease disappears if $R_0<1$, which means that the disease-free equilibrium is globally asymptotically stable. The system is not destabilized by the delay, which leads to periodic oscillations. The numerical simulations validate the theoretical analysis, which illustrates the importance of delay and disinfection in cholera prevention and control.

Author Keywords: Basic reproduction number, Cholera, Delay differential equations, Disinfection, Lyapunov function, Stability analysis

This thesis is focused on studying the population dynamics of a predator-prey system in a patchy environment, taking anti-predation responses into consideration. Firstly, we conduct mathematical analysis on the equilibrium solutions of the system. Using techniques from calculus we show that particular steady state solutions exist when the parameters of the system meet certain criteria. We then show that a further set of conditions leads to the local stability of these solutions. The second step is to extend the existing mathematical analysis by way of numerical simulations. We use octave to confirm the previous results, as well as to show that more complicated dynamics can exist, such as stable oscillations. We consider more complex and meaningful functions for nonlinear dispersal between patches and nonlinear predation, and show that the proposed model exhibits behaviours we expect to see in a population model.

Author Keywords: Anti-predation response, Asymptotic stability, Dispersal, Patch model, Population dynamics, Predator-prey

Recent developments in Artificial Intelligence (AI), particularly Large Language Models (LLMs), have provided powerful tools for Natural Language Processing (NLP) tasks like sentiment analysis. However, their fine-tuning and deployment present challenges, specifically in terms of computational efficiency and high training costs. To address these challenges, this work applies optimization techniques such as Quantized Low-Rank Adaptation (QLoRA) for parameter-efficient fine-tuning, followed by Generalized Post-Training Quantization (GPTQ) on the Llama 3.1 LLM. To evaluate these optimizations, we apply the model to a practical task: hate speech detection, using a curated dataset comprising of X (formerly Twitter) posts. Overall, the optimized model achieved a 67% reduction in size along with significant improvements in classification accuracy and inference speed compared to the base model.

Author Keywords: Generalized Post-Training Quantization, Large Language Models, Low-Rank Adaptation, Parameter-Efficient Fine-Tuning, Quantized Low-Rank Adaptation

The spectral properties of time series data reveal underlying processes but require complete datasets, often unavailable due to missing values and irregular sampling.This thesis uses a computational simulations framework to evaluate the perfor- mance of the Hybrid Wiener Interpolator [3], a novel method designed to reconstruct nonstationary time series data, thus making said data amenable for spectrum analysis. This research evaluates the Hybrid Wiener Interpolator's ability to handle nonstation- ary data and data gaps, comparing its performance to other interpolation methods under different stationarity and data integrity conditions. The results illuminate the robustness of this interpolator in scenarios typical of scientific datasets, offering a promising approach for enhancing spectrum estimation in the presence of non-ideal data conditions

Author Keywords: ARIMA Models, Data Imputation, Interpolation, Stationarity, Time Series, Time Series Simulations

This research presented an AI-driven methodology for precise COVID-19 detectionin medical diagnostic imaging using chest X-ray images. The primary focus was on developing an optimized model using convolutional neural networks (CNNs) and leveraging transfer learning as a low-level feature extractor method. Significant attention was also so given to data transformation and enhancement techniques to improve the information content and distinguishability of non-linearities. The proposed methodology enabled the flexibility to apply multiple models within the framework and identify the most suitable model for the specific task at hand. By emphasizing state-of-the-art CNN models and employing a strategic exploration-exploitation trade-off, this study identified a robust model with heightened accuracy. The results demonstrated a model accuracy of 90.11%, a sensitivity of 91.16%, and a precision of 89.19%, highlighting the model's effectiveness in accurately identifying both true positive and true negative test cases.

Author Keywords: Automatic Bayesian Optimization, Convolution Neural Networks, COVID- 19 Detection, Deep Learning, Medical Diagnostic Imaging, Transfer Learning

This thesis develops an Long Short-Term Memory (LSTM) neural network model to assess the relationship between ambient air pollutant exposure and public health risks, accommodating both linear and nonlinear associations with distributed lags.The research makes three key contributions. First, Maximal Information Coefficient (MIC) methods identify the most relevant air pollutants and their associations with health outcomes. Second, an LSTM model extracts temporally dependent features from exposure series to estimate health impacts. Finally, the model's potential in air pollution epidemiology is explored using Local Interpretable Model-Agnostic Explanations (LIME) to interpret the exposure-health response relationship.

Author Keywords: air pollution epidemiology, Deep Learning, explainable AI, Long Short-Term Memory, Maximal Information Coefficient, public health assessmen

This thesis presents a comprehensive investigation into machine learning approaches forlandmine detection using thermal imagery. It addresses both classification and precise lo- calization challenges that are integral for humanitarian demining operations. The research encompasses two complementary methodological frameworks: comparative evaluation of traditional machine learning versus deep learning approaches, then followed by an imple- mentation of hyperparameter optimization for enhanced safety performance. The foundation of the study demonstrates that traditional machine learning methods achieve competitive classification performance. Conventional models achieved significant performance with the Random Forest and RestNet50 respectively scoring accuracies of 91.88% and 94.29%, though struggle to achieve >10% when tasked with classification and localization. Expanding on this foundation, we addresses this gap through multi-task learn- ing frameworks that simultaneously optimize for both detection and precise localization. Through systematic hyperparameter tuning across 64 configurations, the optimized multi- task approach achieves 90% detection accuracy with 92% precision while providing precise bounding box localization, representing a 37.5% reduction in false negatives. These find- ings demonstrate that while traditional machine learning offers computational efficiency for basic detection, multi-task deep learning frameworks provide significant performance gains when requiring precise spatial localization, which is an important requirement in demining operations.

Author Keywords: computer-vision, demining, humanitarian, Landmines, Multi-Task Learning, YOLO

This thesis introduces a weather-driven bio-economic optimization model for agricultural planning and decision-making. The model integrates weather simulations—including precipitation, temperature, relative humidity, and reference evapotranspiration (ETo)—to estimate crop yields using the AquaCrop simulator. These yield estimates are then incorporated into a multi-objective optimization (MOO) model that aims to maximize gross profit and economic water productivity (ET), while minimizing land use. The MOO model's results provide insights into key agricultural planning questions, such as what, where, when, and how much to plant.The findings demonstrate the model's potential to enhance agricultural decision-making by offering optimized crop combinations that improve both economic returns and land use efficiency. This research contributes to the development of a dynamic agricultural planning model by integrating weather forecasting, crop simulation, and multi-objective optimization.

Author Keywords: AquaCrop, Artificial neural network, Markov chains, Multi-objective optimization, Reference evapotranspiration, Stochastic differential equation