Phobias, characterized by irrational fears of specific objects or situations, can profoundly affect an individual’s quality of life. This research presents a comprehensive investigation into phobia classification, where we propose a novel dataset of 811,569 English tweets from user timelines spanning 102 phobia subtypes over six months, including 47,614 self-diagnosed phobia users. BERT models were leveraged to differentiate non-phobia from phobia users and classify them into 65 specific phobia subtypes. The study produced promising results, with the highest f1-score of 78.44% in binary classification (phobic user or not phobic user) and 24.01% in a multi-class classification (detecting the specific phobia subtype of a user). This research provides insights into people with phobias on social media and emphasizes the capacity of natural language processing and machine learning to automate the evaluation and support of mental health.

Obesity, a complex condition involving genetic, behavioral, socioeconomic, and environmental factors, poses significant health risks and contributes to increased morbidity and mortality. Bariatric surgery is an effective treatment for individuals with severe obesity, resulting in substantial weight loss. However, weight regain remains a significant challenge in the long-term after surgery. This study focuses on analyzing movement patterns of patients who have undergone bariatric surgery using wearable accelerometry to investigate the relationships between movement behaviors, body composition, and weight regain. An intelligent system employing machine learning techniques was utilized to predict total fat percentage and visceral fat. Results indicate that Multivariate Adaptive Regression Splines and Gradient Boosting models show promising performance in predicting fat percentage and visceral fat. Furthermore, the study reveals associations between age, sedentary behavior, post-surgery BMI, day-and night-time movement, and body composition following bariatric surgery. These findings contribute to a better understanding of factors influencing weight regain and may inform future interventions to promote long-term weight loss maintenance in bariatric surgery patients.

Infectious diseases have a profound impact on human society, and although they can cause serious consequences, including loss of life and economic impacts, vaccinations can prevent or minimize their impact. Vaccination strategies, including applying limited vaccines to a population to minimize outbreaks, are crucial to maximize vaccination effectiveness. An evolutionary computation system is designed for generating vaccination strategies based on a social contact network’s topology, as well as individual’s age and comorbidity characteristics, to examine how age and comorbidities impact vaccination strategies. After testing the candidate strategies on multiple graphs and analyzing the results, the results indicated that the base strategies (e.g. vaccinating high degree nodes, vaccinating the population randomly) perform the worst when minimizing the maximum and total number of infected, while normal strategies (derived from general Power-Law Cluster Graphs) were effective, and aged-derived strategies were more effective than comorbidity-derived strategies. It was observed that by implementing these attributes in the graph topology, rather than considering them as graph measures, the effectiveness of the regular strategies could be increased.

Obesity is defined as a body mass index (BMI) of 30 kg/m2or greater. Obesity is a complex medical condition characterized by excess body fat that increases the risk of developing co-morbidities like cardiovascular disease (CVD) and diabetes. Excess weight is a global concern with an ever-growing prevalence. This research study aims to develop effective and practical post-surgical guidelines based on clinical data to determine potential predictors of surgical success. With this information, clinicians and patients can work together to minimize long-term weight regain post-surgery. To fulfill this goal, we applied various machine learning (ML) algorithms to model our dataset and provide a list of critical features to target to have more successful long-term weight-loss outcomes. Fortunately, our ML models achieved outstanding Mean Absolute Errors (0.10 - 0.22).

A generative evolutionary algorithm is used to create personal contact networks representing which individuals can infect others during an infectious disease scenario. Two problems are considered: (i) finding networks that maximize the length of a simulated epidemic, and (ii) finding networks that match given epidemic profiles. A significant innovation is the introduction of weighted edges to represent the strength of the contact between individuals. Different weight initialization conditions are investigated and evaluated for their performance using a parameter selection mechanism designed to explore the parameter space. Results show that weighted edges were able to increase the overall performance achieved by the evolved networks for both problems considered. Furthermore, it is shown that initializing the weights with a value greater than one further improves performance. The results of the parameter selection mechanism were used to test additional parameter settings thoroughly which further maximize the length of the simulated epidemic for the evolved graphs.

How best to apply vaccines to a population is an open problem. It is trivial to derive intuitive strategies, but until tested, their efficacy is not known. This problem is particularly challenging when considering the dynamics of social contact networks and their changes over time. A system for automatically discovering tested vaccination strategies with evolutionary computation has been improved upon to include additional graph metrics and to generate vaccination strategies for dynamic graphs, something that is expected of real social networks within communities. The system's ability to generate effective strategies was demonstrated along with a comparison of the strategies developed when fit to a static graph versus a dynamic graph. It was observed that the additional computational resources required to generate strategies on a dynamic graph may not be necessary as strategies developed for static graphs performed similarly well; however, the authors are careful to acknowledge that results may differ significantly when adjusting the systems many parameters.

Necrotizing Fasciitis (NF), or Necrotizing Soft-Tissue Infection (NSTI), is a rare infection that poses a significant threat to health. In the absence of a proper diagnosis, the infection can spread rapidly causing extensive tissue necrosis and death - mortality rate of 20% - 35%. Due to inadequate resources, little progress has been made for the automatic detection of NF. We have prepared a novel dataset containing images of affected human organs by NF using an internet image search. The dataset contains 693 images in total, containing raw, augmented, and non-NF images. A system has been developed for performing automated detection of NF with an Artificial Neural Network. We have evaluated the YOLOv3 object recognition model for five arrangements of our dataset and compared the performance for these different data arrangements after running each five times. The datasets were split into 80% train data and 20% test data, and for performance measures, we have taken into account the evaluation metrics: Intersection over Union (IoU) and Average Precision (AP). We obtained the highest average AP score of 57.97% for the dataset with raw data and augmentation and the highest average IoU score of 61.94% for dataset with raw data, augmentation, and negative images. The initial finding of this work can be further improved and become a substantial contribution to clinical arrangements for the diagnosis and management of NF.

The COVID-19 pandemic, caused by the SARS-CoV-2 virus, led to a global health crisis, with more than 157 million cases confirmed infected by May 2021. Effective medication is desperately needed. Predicting drug-target interaction (DTI) is an important step to discover novel uses of chemical structures. Here, we develop a pipeline to predict novel DTIs based on the proteins of the coronavirus. Different datasets (human/SARS-CoV-2 Protein-Protein interaction (PPI), Drug-Drug similarity (DD sim), and DTIs) are used and combined. After mapping all datasets onto a heterogeneous graph, path-related features are extracted. We then applied various machine learning (ML) algorithms to model our dataset and predict novel DTIs among unlabeled pairs. Possible drugs identified by the models with a high frequency are reported. In addition, evidence of the efficiency of the predicted medicines by the models against COVID-19 are presented. The proposed model can then be generalized to contain other features that provide a context to predict medicine for different diseases.

Parkinson's disease is a neurodegenerative disease that affects close to 10 million with various symptoms including tremors and changes in gait. Observing differences or changes in an individual's manifestations of gait may provide a mechanism to identify Parkinson's disease and understand specific changes. In this study, timeseries data from both Control subjects and Parkinson's disease patients was modelled with symbolic regression and extreme gradient boosting. Model effectiveness was analyzed along with the differences in the models between modelling strategies, between Control subjects and Parkinson's disease patients, and between normal walking and walking while under a cognitive load. Both modelling strategies were found to effective. The symbolic regression models were more easily interpreted, while extreme gradient boosting had higher overall accuracy. Interpretation of the models identified certain characteristics that distinguished Control subjects from Parkinson's disease patients and normal walking conditions from walking while under a cognitive load.

Evolutionary algorithms are used to generate personal contact networks, modelling human populations, that are most likely to match a given epidemic profile. The Susceptible-Infected-Removed (SIR) model is used and also expanded upon to allow for an extended period of infection, termed the SIIR model. The networks generated for each of these models are thoroughly evaluated for their ability to match nine different epidemic profiles. The addition of the SIIR model showed that the model of infection has an impact on the networks generated. For the SIR and SIIR models, these differences were relatively minor in most cases.

It is important to understand how best to apply a limited number of vaccines to a population such that the spread of a disease, like SARS-CoV-2, is minimized. Although intuition provides a number of mitigation strategies that may be effective, they remain largely untested.A system was developed to test a given disease mitigation strategy. It is designed to work with a graph representing a real social network. A Genetic Programming system was used to discover novel mitigation strategies that are easily interpretable by a public health decision maker.Effective strategies were developed by the GP system. The strategies are easily explainable and intuitive. Novel mitigation strategies were compared to simple baseline strategies with varying success using a number of different metrics. Many of these strategies proved effective in general, however the topology of the graph influences the effectiveness of a strategy.The system has been made publicly available and the authors call on the research community to contribute their own mitigation strategies and measure their efficacy.

Traditionally, models developed to estimate resting energy expenditure (REE) in the bariatric population have been limited to linear modelling based on data from `normal' or `overweight' individuals - not `obese'. This type of modelling can be restrictive and yield functions which poorly estimate this important physiological outcome.Linear and nonlinear models of REE for individuals after bariatric surgery are developed with linear regression and symbolic regression via genetic programming. Features not traditionally used in REE modelling were also incorporated and analyzed and genetic programming's intrinsic feature selection was used as a measure of feature importance.A collection of effective new linear and nonlinear models were generated. The linear models generated outperformed the nonlinear on testing data, although the nonlinear models fit the training data better. Ultimately, the newly developed linear models showed an improvement over existing models and the feature importance analysis suggested that the typically used features (age, weight, and height) were the most important.

Parkinson's Disease is a disorder with diagnostic symptoms that include a change to a walking gait. The disease is problematic to diagnose. An objective method of monitoring the gait of a patient is required to ensure the effectiveness of diagnosis and treatments. We examine the suitability of Extreme Gradient Boosting (XGBoost) and Artificial Neural Network (ANN) Models compared to Symbolic Regression (SR) using genetic programming that was demonstrated to be successful in previous works on gait. The XGBoost and ANN models are found to out-perform SR, but the SR model is more human explainable.

Smart devices are becoming more universally adopted and can be used to track and model user activity and monitor for abnormalities. Deviations from what is expected may indicate that a fall is imminent or that an injury has been sustained. Healthcare practitioners can use descriptive models of human kinematics as a tool to monitor patient recovery. This work extends previous work which generated descriptive nonlinear symbolic models of human kinematics with genetic programming. Previously, linear models were developed and compared to the nonlinear models. Although the linear models fit the data well, they were significantly worse than the nonlinear models. In this phase of the project, ensembles of nonlinear models were created to more accurately fit and classify data. Different model selection strategies for the ensembles were investigated. As one would expect, ensembles of models were significantly better than a single model classifier. It was also observed that, although more models in the ensemble yielded better results, only 2 models were required to obtain significantly better results. It was also observed that a random model selection strategy for the ensembles produced competitive results when compared to a more rigorous model selection strategy.

The brain is a nonlinear computational system; however, most methods employed in finding functional connectivity models with functional magnetic resonance imaging (fMRI) data produce strictly linear models - models incapable of truly describing the underlying system.Genetic programming is used to develop nonlinear models of functional connectivity from fMRI data. The study builds on previous work and observes that nonlinear models contain relationships not found by traditional linear methods. When compared to linear models, the nonlinear models contained fewer regions of interest and were never significantly worse when applied to data the models were fit to. Nonlinear models could generalize to unseen data from the same subject better than traditional linear models (intrasubject). Nonlinear models could not generalize to unseen data recorded from other subjects (intersubject) as well as the linear models, and reasons for this are discussed. This study presents the problem that many, manifestly different models in both operators and features, can effectively describe the system with acceptable metrics.

Parkinson's disease (PD) is a degenerative disorder of the central nervous system that has many debilitating symptoms which affect the patient's motor system and can cause significant changes in their gait. By using genetic programming, we aim to develop descriptive symbolic nonlinear models of PD patient gait from time series data recorded from pressure sensors under subjects' feet. When compared to popular types of linear regression (OLS and LASSO), the nonlinear models fit their data better and generalize to unseen data significantly better. It was found that models developed for healthy control subjects generalized to other control subjects well, however the models trained on subjects with PD did not generalize well to other PD patients, which complicates the issue of being able to detect the progression of the disease. It is suspected that health care professionals can have difficulty classifying PD due to a lack of accurate data from patient reports; having individually trained models for active monitoring of patients would help in effectively diagnosing PD.

Side Effect Machines (SEMs) are an extension of finite state machines which place a counter on each node that is incremented when that node is visited. Previous studies examined a genetic algorithm to discover node connections in SEMs for edit metric decoding for biological applications, namely to handle sequencing errors. Edit metric codes, while useful for decoding such biologically created errors, have a structure which significantly differentiates them from other codes based on Hamming distance. Further, the inclusion of biologically- motivated restrictions on allowed words makes development of decoders a bespoke process based on the exact code used. This study examines the use of evolutionary programming for the creation of such decoders, thus allowing for the number of states to be evolved directly, not witnessed in previous approaches which used genetic algorithms. Both direct and fuzzy decoding are used, obtaining correct decoding rates of up to 95% in some SEMs.

Smart devices are becoming an extension of ourselves that contain sensitive information and are often targeted for theft. The development of an intelligent and reliable means of user identification and authentication is critical. Not only can the development of user models performing tasks be used for user and task identification, but systems can also notify individuals if there is a potential health concern. The construction of an idealized model of human locomotion may give medical care providers a better understanding of individual differences and guide therapy and treatment. Data was gathered from a smartwatch worn by six subjects performing five different tasks and Genetic Programming was used to perform symbolic regression - a model free, nonlinear type of regression analysis. Symbolic regression was applied to smartwatch data and a collection of nonlinear closed form symbolic mathematical models were generated. Not only did these models fit the data well, but they provided insight into the underlying system. With only 5 seconds of unseen data, the models could classify which subjects were performing which task with 83.9% accuracy when chance was only 3.33%.

In contrast to conventional, univariate analysis, various types of multivariate analysis have been applied to functional magnetic resonance imaging (fMRI) data. In this paper, we compare two contemporary approaches for multivariate regression on task-based fMRI data: linear regression with ridge regularization and non-linear symbolic regression using genetic programming. The data for this project is representative of a contemporary fMRI experimental design for visual stimuli. Linear and non-linear models were generated for 10 subjects, with another 4 withheld for validation. Model quality is evaluated by comparing R scores (Pearson product-moment correlation) in various contexts, including single run self-fit, within-subject generalization, and between-subject generalization. Propensity for modelling strategies to overfit is estimated using a separate resting state scan. Results suggest that neither method is objectively or inherently better than the other.

Patients who suffered a traumatic brain injury (TBI) require special care, and physicians often monitor intercranial pressure (ICP) as it can greatly aid in management. Although monitoring ICP can be critical, it requires neurosurgery, which presents additional significant risk. Monitoring ICP also aids in clinical situations beyond TBI, however the risk of neurosurgery can prevent physicians from gathering the data. The need for surgery may be eliminated if ICP could be accurately inferred using noninvasive physiological measures. Genetic programming (GP) and linear regression were used to develop nonlinear and linear mathematical models describing the relationships between intercranial pressure and a collection of physiological measurements from noninvasive instruments. Nonlinear models of ICP were generated that not only fit the subjects they were trained on, but generalized well across other subjects. The nonlinear models were analysed and provided insight into the studied underlying system which led to the creation of additional models. The new models were developed with a refined search, and were more accurate and general. It was also found that the relations between the features could be explained effectively with a simple linear model after GP refined the search.

Despite significant effort, there is currently no formal or de facto standard framework or format for constructing, representing, or manipulating general neural networks. In computational neuroscience, there have been some attempts to formalize connectionist notations and generative operations for neural networks, including Connection Set Algebra, but none are truly formal or general. In computational intelligence (CI), though the use of linear algebra and tensor-based models are widespread, graph-based frameworks are also popular and there is a lack of tools supporting the transfer of information between systems. To address these gaps, we exploited existing results about the connection between linear and relation algebras to define a concise, formal algebraic framework that generalizes graph and tensor-based neural networks. For simplicity and compatibility, this framework is purposefully defined as a minimal extension to linear algebra. We demonstrate the merits of this approach first by defining new operations for network composition along with proofs of their most important properties. An implementation of the algebraic framework is presented and applied to create an instance of an artificial neural network that is compatible with both graph and tensor based CI frameworks. The result is an algebraic framework for neural networks that generalizes the formats used in at least two systems, together with an example implementation.

The vast majority of methods employed in the analysis of functional Magnetic Resonance Imaging (fMRI) produce exclusively linear models; however, it is clear that linear models cannot fully describe a system with the observed behavioral complexity of the human brain --- an intrinsically nonlinear system. By using tools embracing the possibility of modeling the underlying nonlinear system we may uncover meaningful undiscovered relationships which further our understanding of the brain.

The idea of using the gait of a walking person as a biometric identification method has been seen in a number of proposed authentication methods, yet previous works focus on the addition of other authentication methods along with the gait, or require a stationary sensor attached to the hip of the user. This paper uses Genetic Programming to model an identification gait fingerprint for two users, whose walking data was recorded from the accelerometer in a commercially available phone. With the phone freely placed within a pocket, users moved without a fixed protocol at a normal, nonuniform pace. This design of data collection more closely matches the real world applications of such a method. The highly specialized Genetic Programming system with multiple modular enhancements was implemented to perform symbolic regression. The system was demonstrated to be robust to noise and was able to effectively model each dataset with high accuracy. It was also determined that a model could be generated for a subject's whole dataset from only a single step's worth of data. Top models were applied to other subject's data in order to evaluate the uniqueness of these mathematical models.

The Fragment Assembly Problem is a major component of the DNA sequencing process that is identified as being NP-Hard. A variety of approaches to this problem have been used, including overlap-layout-consensus, de Bruijn graphs, and greedy graph based algorithms. The overlap-layout-consensus approach is one of the more popular strategies which has been studied on a collection of heuristics and metaheuristics. In this study heuristics and Genetic Algorithm variations are combined to exploit their respective benefits. These algorithms were able to produce results that surpassed the best results obtained by a collection of state-of-the-art metaheuristics on ten of sixteen popular benchmark data sets.

Quaternary error-correcting codes defined over the edit metric may be used as labels to track the origin of sequence data. When used in such applications there are typically additional restrictions that are biologically motivated, such as a required GC content or the avoidance of certain patterns. As a result such codes can not be expected to have a regular structure, making decoding particularly challenging. Previous work on decoding edit codes considered the use of side effect machines for decoding, successfully decoding up to 93.86% of error vectors. In this study the recentering/restarting algorithm is used in combination with side effect machines and an alternative representation based upon transpositions. Using the same data as in the previous work, the rate of successful decoding was significantly improved, with many cases obtaining rates very close to 100%.