Alzheimer's Disease (AD) poses significant challenges in research due to limited access to longitudinal patient data caused by privacy constraints. This study uses deep learning, specifically Variational Autoencoders (VAEs), to generate synthetic datasets that replicate real-world AD data while preserving privacy. These datasets were used to train and compare survival analysis models, including Survival XGBoost, Survival Transformers, and the Cox Proportional Hazards model. Results show that deep learning and boosting models outperform the standard Cox model in predicting the progression of AD, achieving a higher degree of accuracy. Our findings suggest that combining synthetic data with advanced machine learning models can improve predictive capabilities in healthcare research, particularly when real patient data is restricted.

In recent years, attention within the clinical prediction community has turned to the use of survival machine learning as a tool for predicting the risk of developing a disease as a function of time. The current work seeks to contribute to existing literature which demonstrates the utility of these methods when applied to a dementia prediction context. We use the Alzheimer's Disease Neuroimaging Initiative ADNI dataset and model deterioration within two distinct groups, those deemed cognitively normal and those with a formal diagnosis of Mild Cognitive Impairment. In agreement with existing literature we find that survival machine learning outperforms standard survival analysis methods such as Cox PH model, and has very good predictive ability. We propose an innovative approach to predicting dementia diagnosis risk on ADNI, which explores the use of survival neural network and survival extreme gradient boosting techniques that have hitherto seldom been applied to this context. The stability of our models was investigated within a Monte Carlo simulation framework.

The Multiple Traveling Salesman Problem (multiple-TSP) is a straightforward extension of the well-known Traveling Salesman Problem (TSP), in which more salesmen must visit a set of interconnected cities. Ant Colony Optimization (ACO) algorithms are designed to build sequentially the solutions, aspect which on multiple-TSP imposes new challenges. Compared to TSP which deals with one sample space - the set of cities (locations), multiple-TSP involves two sample spaces: the set of salesmen (agents) and the set of cities. Existing ACO algorithms addressing multiple-TSP are two-phase sampling procedures which, firstly, independently sample from the first set (the set of salesmen), and then conditionally sample from the second set. Our claim is that a joint sampling mechanism, which will exploit a joint probability space, is likely to lead to superior results. We validate our hypothesis by implementing five ACO-based algorithms to solve multiple-TSP: three of them are two-phase algorithms exploring various methods to sample from the salesmen space, while two of them implement the joint sampling scheme. The results are analyzed both in a single-objective manner that considers the minimization of the longest tour, and also from a bi-objective perspective that considers two conflicting objectives: 1) minimization of the total traveled distance and 2) work balancing – which amounts to minimizing the amplitude of the costs of individual tours.

Worldwide, it is forecasted that 131.5 million people will suffer from dementia by 2050, and the annual cost of care will increase from 818 billion USD in 2016 to 2 trillion USD by 2030, with burgeoning social consequences. Given a timely prediction of a dementia outcome in patients, appropriate mitigating interventions can be applied to reduce risk. However such prediction facilities need to be made available to wider populations, and these facilities cannot rely on specialised, costly and invasive testing (such as neuroimaging, cerebrospinal fluid collection, etc which constitute important instruments used in diagnosis), for interventions to have a meaningful quantitative impact. Hence an emerging need exists for the wider application of prognostic measures which can be deployed using lower cost data sources such as longitudinal records routinely collected by general practices. This paper proposes an efficient prediction modelling approach to the risk of dementia, using CPRD data collected from GP practices in UK, and based on machine learning in particular the Gradient Boosting Machines model combined with a survival model such as the Cox Proportional Hazard, encapsulated in a semi-supervised learning and model calibration methodology.

This paper explores deterioration in Alzheimer’s Disease using Machine Learning. Subjects were split into two datasets based on baseline diagnosis (Cognitively Normal, Mild Cognitive Impairment), with outcome of deterioration at final visit (a binomial essentially yes/no categorisation) using data from the Alzheimer’s Disease Neuroimaging Initiative (demographics, genetics, CSF, imaging, and neuropsychological testing etc). Six machine learning models, including gradient boosting, were built, and evaluated on these datasets using a nested cross-validation procedure, with the best performing models being put through repeated nested cross-validation at 100 iterations. We were able to demonstrate good predictive ability using CART predicting which of those in the cognitively normal group deteriorated and received a worse diagnosis (AUC = 0.88). For the mild cognitive impairment group, we were able to achieve good predictive ability for deterioration with Elastic Net (AUC = 0.76).

The special session "Data Science is Computational Psychiatry and Psychiatric Research" at the 5th IEEE International Conference in Data Science and Advanced Analytics in Turin, Italy 2018 presents papers specifically addressing psychiatric research. In this overview, we describe the challenges of psychiatric research and demonstrates how the presented papers approach some of the problems.

Dementia is one of the most feared illnesses that has a growing year-to-year negative global impact, having a health and social care cost higher than cancer, stroke and chronic heart disease, taken together. Without the availability of a cure, nor a standardised clinical test, the utilisation of machine learning methods to identify individuals that are at risk of developing dementia could bring a new step towards proactive intervention. This study's goal is to carry out a precursor analysis leading to building classification models with enhanced capabilities for differentiating diagnoses of CN (Cognitively Normal), MCI (Mild Cognitive Impairment) and Dementia. The predictive modelling approach we propose is based on the ReliefF method combined with statistical permutation tests for feature selection, and on model training, tuning, and testing based on algorithms such as Random Forests, Support Vector Machines, Gaussian Processes, Stochastic Gradient Boosting, and eXtreme Gradient Boosting. Stability of model performances were studied in computationally intensive Monte Carlo simulations. The results consistently show that our models accurately detect dementia, and also mild cognitive impairment patients by only using the inclusion of baseline measurements as predictors, thus illustrating the importance of baseline measurements. The best results issued from Monte Carlo were achieved by eXtreme Gradient Boosting optimised models, with an accuracy of 0.88 (SD 0.02), a sensitivity of 0.93 (SD 0.02) and a specificity of 0.94 (SD 0.01) for dementia, and a sensitivity of 0.86 (SD 0.02) and a specificity of 0.9 (SD 0.02) for mild cognitive impairment. These results support in particular future developments for a risk-based method that can identify an individual's risk of developing dementia.

Wheezing is common among children and ~50% of those under 6 years of age are thought to experience at least one episode of wheeze. However, due to the heterogeneity of symptoms there are difficulties in treating and diagnosing these children. `Phenotype specific therapy' is one possible avenue of treatment, whereby we use significant pathology and physiology to identify and treat pre-schoolers with wheeze. By performing feature selection algorithms and predictive modelling techniques, this study will attempt to determine if it is possible to robustly distinguish patient diagnostic categories among pre-school children. Univariate feature analysis identified more objective variables and recursive feature elimination a larger number of subjective variables as important in distinguishing between patient categories. Predicative modelling saw a drop in performance when subjective variables were removed from analysis, indicating that these variables are important in distinguishing wheeze classes. We achieved 90%+ performance in AUC, sensitivity, specificity, and accuracy, and 80%+ in kappa statistic, in distinguishing ill from healthy patients. Developed in a synergistic statistical - machine learning approach, our methodologies propose also a novel ROC Cross Evaluation method for model post-processing and evaluation. Our predictive modelling's stability was assessed in computationally intensive Monte Carlo simulations.

Smart phones have become ubiquitous in the recent years, which opened up a new opportunity for rediscovering the Experience Sampling Method (ESM) in a new efficient form using mobile apps, and provides great prospects to become a low cost and high impact mHealth tool for psychiatry practice. The method is used to collect longitudinal data of participants' daily life experiences, and is ideal to capture fluctuations in emotions (momentary mental states) as an early indicator for later mental health disorder. In this study ESM data of patients with psychosis and controls were used to examine emotion changes and identify patterns. This paper attempts to determine whether aggregated ESM data, in which statistical measures represent the distribution and dynamics of the original data, are able to distinguish patients from controls. Variable importance, recursive feature elimination and ReliefF methods were used for feature selection. Model training and tuning, and testing were performed in nested cross-validation, and were based on algorithms such as Random Forests, Support Vector Machines, Gaussian Processes, Logistic Regression and Neural Networks. ROC analysis was used to post-process these models. Stability of model performances was studied using Monte Carlo simulations. The results provide evidence that pattern in mood changes can be captured with the combination of techniques used. The best results were achieved by SVM with radial kernel, where the best model performed with 82% accuracy and 82% sensitivity.

Over the last two decades, a significant body of research has established a link between cannabis use and psychotic outcomes. In this study, we aim to propose a novel symbiotic machine learning and statistical approach to pattern detection and to developing predictive models for the onset of first-episode psychosis. The data used has been gathered from real cases in cooperation with a medical research institution, and comprises a wide set of variables including demographic, drug-related, as well as several variables specifically related to the cannabis use. Our approach is built upon several machine learning techniques whose predictive models have been optimised in a computationally intensive framework. The ability of these models to predict first-episode psychosis has been extensively tested through large scale Monte Carlo simulations. Our results show that Boosted Classification Trees outperform other models in this context, and have significant predictive ability despite a large number of missing values in the data. Furthermore, we extended our approach by further investigating how different patterns of cannabis use relate to new cases of psychosis, via association analysis and Bayesian techniques.

The frequent ups and downs are characteristic to the stock market. The conventional standard models that assume that investors act rationally have not been able to capture the irregularities in the stock market patterns for years. As a result, behavioural finance is embraced to attempt to correct these model shortcomings by adding some factors to capture sentimental contagion which may be at play in determining the stock market. This paper assesses the predictive influence of sentiment on the stock market returns by using a non-parametric nonlinear approach that corrects specific limitations encountered in previous related work. In addition, the paper proposes a new approach to developing stock market volatility predictive models by incorporating a hybrid GARCH and artificial neural network framework, and proves the advantage of this framework over a GARCH only based framework. Our results reveal also that past volatility and positive sentiment appear to have strong predictive power over future volatility.

We address the problem of defining semantics for logic programs in presence of incomplete and contradictory information coming from different sources. The information consists of facts that a central server collects and tries to combine using (a) a set of logical rules, that is, a logic program, and (b) a hypothesis representing the server's own estimates. In such a setting incomplete information from a source or contradictory information from different sources necessitate the use of many-valued logics in which programs can be evaluated and hypotheses can be tested. To carry out such activities we propose a formal framework based on bilattices such as Belnap's four-valued logics. In this framework we work with the class of programs defined by Fitting and we propose hypothesis-based semantics for such programs. We also establish an intuitively appealing connection between our hypothesis testing mechanism, on the one hand, and the well-founded semantics and Kripke-Kleene semantics of Datalog programs with negation, on the other hand.

Answer-perturbation techniques for the protection of statistical databases have been previously introduced (Luchian and Stamate, 1992); they are flexible (perturbation kept under control), modular (do not interact with the DBMS) techniques, which compare favorably to previous protection techniques. In this paper, we generalise the answer-perturbation techniques w.r.t. the operation used for modifying the exact answers (thus enhancing the level of protection). Experimental results are also included; they indicate statistical soundness of our techniques.>