There has been an increasing interest in survival analysis with interval-censored data, where the event of interest (such as infection with a disease) is not observed exactly but only known to happen between two examination times. However, because so much research has been focused on right-censored data, so many statistical tests and techniques are available for right-censoring methods, hence interval-censoring methods are not as abundant as those for right-censored data. In this study, right-censoring methods are used to fit a proportional hazards model to some interval-censored data. Transformation of the interval-censored observations was done using a method called mid-point imputation, a method which assumes that an event occurs at some midpoint of its recorded interval. Results obtained gave conservative regression estimates but a comparison with the conventional methods showed that the estimates were not significantly different. However, the censoring mechanism and interval lengths should be given serious consideration before deciding on using mid-point imputation on interval-censored data.

With the increasing amount of collected data, large-scale high-dimensional data analysis is becoming essential in many areas. These data can be analyzed either by using fully computational methods or by leveraging human capabilities via interactive visualization. However, each method has its drawbacks. While a fully computational method can deal with large amounts of data, it lacks depth in its understanding of the data, which is critical to the analysis. With the interactive visualization method, the user can give a deeper insight on the data but suffers when large amounts of data need to be analyzed. Even with an apparent need for these two approaches to be integrated, little progress has been made. As ways to tackle this problem, computational methods have to be re-designed both theoretically and algorithmically, and the visual analytics system has to expose these computational methods to users so that they can choose the proper algorithms and settings. To achieve an appropriate integration between computational methods and visual analytics, the thesis focuses on essential computational methods for visualization, such as dimension reduction and clustering, and it presents fundamental development of computational methods as well as visual analytic systems involving newly developed methods. The contributions of the thesis include (1) the two-stage dimension reduction framework that better handles significant information loss in visualization of high-dimensional data, (2) efficient parametric updating of computational methods for fast and smooth user interactions, and (3) an iteration-wise integration framework of computational methods in real-time visual analytics. The latter parts of the thesis focus on the development of visual analytics systems involving the presented computational methods, such as (1) Testbed: an interactive visual testbed system for various dimension reduction and clustering methods, (2) iVisClassifier: an interactive visual classification system using supervised dimension reduction, and (3) VisIRR: an interactive visual information retrieval and recommender system for large-scale document data.

Thesis (MScEng)--Stellenbosch University, 2013. / ENGLISH ABSTRACT: Predictive analytics in healthcare has become one of the major focus areas in healthcare delivery worldwide. Due to the massive amount of healthcare data being captured, healthcare providers and health insurers are investing in predictive analytics and its enabling technologies to provide valuable insight into a large variety of healthcare outcomes. One of the latest developments in the field of healthcare predictive modelling (PM) was the launch of the Heritage Health Prize; a competition that challenges individuals from across the world to develop a predictive model that successfully identifies the patients at risk of admission to hospital from a given patient population. The patient-admission predictive algorithm (PAPA) is aimed at reducing the number of unnecessary hospitalisations that needlessly constrain healthcare service delivery worldwide. The aim of the research presented is to determine the feasibility and value of applying PAPAs in the South African healthcare system as part of a preventive care intervention strategy. A preventive care intervention strategy is a term used to describe an out-patient hospital service, aimed at providing preventive care in an effort to avoid unnecessary hospitalisations from occurring. The thesis utilises quantitative and qualitative techniques. This included a review of the current and historic PM applications in healthcare to determine the major expected shortfalls and barriers to implementation of PAPAs, as well as the institutional and operational requirements of these predictive algorithms. The literature study is concluded with a review of the current state of affairs in the South African healthcare system to, firstly, articulate the need for PAPAs and, secondly, to determine whether the public and private sectors provide a suitable platform for implementation (evaluated based on the operational and institutional requirements of PAPAs). Furthermore, a methodology to measure and analyse the potential value-add of a PAPA care intervention strategy was designed and developed. The methodology required a survey of the industry leaders in the private healthcare sector of South Africa to identify, firstly, the current performance foci and, secondly, the factors that compromise the performance of these organisations to deliver high quality, resource-effective care. A quantitative model was developed and applied to an industry leader in the private healthcare sector of South Africa, in order to gauge the resultant impact of a PAPA care intervention strategy on healthcare provider performance. Lastly, in an effort to ensure the seamless implementation and operation of PAPAs, an implementation framework was developed to address the strategic, tactical, and operational challenges of applying predictive analytics and preventive care strategies similar to PAPAs. The research found that the application of PAPAs in the public healthcare sector of South Africa is infeasible. The private healthcare sector, however, was considered a suitable platform to implement PAPAs, as this sector satisfies the institutional and operational requirements of PAPAs. The value-add model found that a PAPA intervention strategy will add significant value to the performance of healthcare providers in the private healthcare sector of South Africa. Noteworthy improvements are expected in the ability of healthcare provider’s to coordinate patient care, patient-practitioner relationships, inventory service levels, and staffing level efficiency and effectiveness. A slight decrease in the financial operating margin, however, was documented. The value-add methodology and implementation support framework provides a suitable platform for future researchers to explore the collaboration of preventive care and PM in an effort to improve healthcare resource management in hospitals. In conclusion, patient-admission predictive algorithms provide improved evidence-based decision making for preventive care intervention strategies. An efficient and effective preventive care intervention strategy improves healthcare provider performance and, therefore, adds significant value to these organisations. With the proper planning and implementation support, the application of PAPA care intervention strategies will change the way healthcare is delivered worldwide. / AFRIKAANSE OPSOMMING: Vooruitskattingsanalises in gesondheidsorg het ontwikkel in een van die mees belangrike fokusareas in die lewering van kwaliteit gesondheidsorg in ontwikkelde lande. Gesondheidsorgverskaffers en lewensversekeraars belê in vooruitskattingsanalise en ooreenstemmende tegnologieë om groot hoeveelhede gesondheidsorg pasiënt-data vas te lê, wat waardevolle insigte bied ten opsigte van ʼn groot verskeidenheid van gesondheidsorg-uitkomstes. Een van die nuutste ontwikkelinge in die veld van gesondheidsorg vooruitskattingsanalises, was die bekendstelling van die “Heritage Health Prize”, 'n kompetisie wat individue regoor die wêreld uitdaag om 'n vooruitskattingsalgoritme te ontwikkel wat pasiënte identifiseer wat hoogs waarskynlik gehospitaliseer gaan word in die volgende jaar en as bron-intensief beskou word as gevolg van die beraamde tyd wat hierdie individue in die hospitaal sal deurbring. Die pasiënt-toelating vooruitskattingsalgoritme (PTVA) het ten doel om onnodige hospitaliserings te identifiseer en te voorkom tem einde verbeterde hulpbronbestuur in gesondheidsorg wêreldwyd te bewerkstellig. Die doel van die hierdie projek is om die uitvoerbaarheid en waarde van die toepassing van PTVAs, as 'n voorkomende sorg intervensiestrategie, in die Suid-Afrikaanse gesondheidsorgstelsel te bepaal. 'n Voorkomende sorg intervensiestrategie poog om onnodige hospitaliserings te verhoed deur die nodige sorgmaatreëls te verskaf aan hoë-riskio pasiënte, sonder om hierdie individue noodwendig te hospitaliseer. Die tesis maak gebruik van kwantitatiewe en kwalitatiewe tegnieke. Dit sluit in 'n hersiening van die huidige en historiese vooruitskattings modelle in die gesondheidsorgsektor om die verwagte struikelblokke in die implementering van PTVAs te identifiseer, asook die institusionele en operasionele vereistes van hierdie vooruitskattingsalgoritmes te bepaal. Die literatuurstudie word afgesluit met 'n oorsig van die huidige stand van sake in die Suid-Afrikaanse gesondheidsorgstelsel om, eerstens, die behoefte vir PTVAs te identifiseer en, tweedens, om te bepaal of die openbare en private sektore 'n geskikte platform vir implementering bied (gebaseer op die operasionele en institusionele vereistes van PTVAs). Verder word 'n metodologie ontwerp en ontwikkel om die potensiële waarde-toevoeging van 'n PTVA sorg intervensiestrategie te bepaal. Die metode vereis 'n steekproef van die industrieleiers in die private gesondheidsorgsektor van Suid-Afrika om die volgende te identifiseer: die huidige hoë-prioriteit sleutel prestasie aanwysers (SPAs), en die faktore wat die prestasie van hierdie organisasies komprimeer om hoë gehalte, hulpbron-effektiewe sorg te lewer. 'n Kwantitatiewe model is ontwikkel en toegepas op een industrieleier in die private Stellenbosch gesondheidsorgsektor van Suid-Afrika, om die gevolglike impak van 'n PTVA sorg intervensiestrategie op prestasieverbetering te meet. Ten slotte, in 'n poging om te verseker dat die implementering en werking van PTVAs glad verloop, is 'n implementeringsraamwerk ontwikkel om die strategiese, taktiese en operasionele uitdagings aan te spreek in die toepassing van vooruitskattings analises en voorkomende sorg strategieë soortgelyk aan PTVAs. Die navorsing het bevind dat die toepassing van PTVAS in die openbare gesondheidsorgsektor van Suid-Afrika nie lewensvatbaar is nie. Die private gesondheidsorgsektor word egter beskou as 'n geskikte platform om PTVAs te implementeer, weens die bevrediging van die institusionele en operasionele vereistes van PTVAs. Die waarde-toevoegings model het bevind dat 'n PTVA intervensiestrategie beduidende waarde kan toevoeg tot die prestasieverbetering van gesondheidsorgverskaffers in die private gesondheidsorgsektor van Suid-Afrika. Die grootste verbetering word in die volgende SPAs verwag; sorg koördinasie, dokter-pasiënt verhoudings, voorraad diensvlakke, en personeel doeltreffendheid en effektiwiteit. 'n Effense afname in die finansiële bedryfsmarge word egter gedokumenteer. 'n Implementering-ondersteuningsraamwerk is ontwikkel in 'n poging om die sleutel strategiese, taktiese en operasionele faktore in die implementering en uitvoering van 'n PTVA sorg intervensiestrategie uit te lig. Die waarde-toevoegings metodologie en implementering ondersteuning raamwerk bied 'n geskikte platform vir toekomstige navorsers om die rol van vooruitskattings modelle in voorkomende sorg te ondersoek, in 'n poging om hulpbronbestuur in hospitale te verbeter. Ten slotte, PTVAs verbeter bewysgebaseerde besluitneming vir voorkomende sorg intervensiestrategieë. 'n Doeltreffende en effektiewe voorkomende sorg intervensiestrategie voeg aansienlike waarde tot die algehele prestasieverbetering van gesondheidsorgverskaffers. Met behoorlike beplanning en ondersteuning met implementering, sal PTVA sorg intervensiestrategieë die manier waarop gesondheidsorg gelewer word, wêreldwyd verander.

Indiana University-Purdue University Indianapolis (IUPUI) / Robustness and generalizability of supervised learning algorithms depend on the quality of the labeled data set in representing the real-life problem. In many real-world domains, however, we may not have full knowledge of the underlying data-generating mechanism, which may even have an evolving nature introducing new classes continually. This constitutes a partially-observed setting, where it would be impractical to obtain a labeled data set exhaustively defined by a fixed set of classes. Traditional supervised learning algorithms, assuming an exhaustive training library, would misclassify a future sample of an unobserved class with probability one, leading to an ill-defined classification problem. Our goal is to address situations where such assumption is violated by a non-exhaustive training library, which is a very realistic yet an overlooked issue in supervised learning. In this dissertation we pursue a new direction for supervised learning by defining self-adjusting models to relax the fixed model assumption imposed on classes and their distributions. We let the model adapt itself to the prospective data by dynamically adding new classes/components as data demand, which in turn gradually make the model more representative of the entire population. In this framework, we first employ suitably chosen nonparametric priors to model class distributions for observed as well as unobserved classes and then, utilize new inference methods to classify samples from observed classes and discover/model novel classes for those from unobserved classes. This thesis presents the initiating steps of an ongoing effort to address one of the most overlooked bottlenecks in supervised learning and indicates the potential for taking new perspectives in some of the most heavily studied areas of machine learning: novelty detection, online class discovery and semi-supervised learning.

Indiana University-Purdue University Indianapolis (IUPUI) / Text mining is a process of extraction of knowledge from unstructured text documents. We have huge volumes of text documents in digital form. It is impossible to manually extract knowledge from these vast texts. Hence, text mining is used to find useful information from text through the identification and exploration of interesting patterns. The objective of this thesis in text mining area is to find compact but high quality frequent patterns from text documents related to neuroscience field. We try to prove that interactive sampling algorithm is efficient in terms of time when compared with exhaustive methods like FP Growth using RapidMiner tool. Instead of mining all frequent patterns, all of which may not be interesting to user, interactive method to mine only desired and interesting patterns is far better approach in terms of utilization of resources. This is especially observed with large number of keywords. In interactive patterns mining, a user gives feedback on whether a pattern is interesting or not. Using Markov Chain Monte Carlo (MCMC) sampling method, frequent patterns are generated in an interactive way. Thesis discusses extraction of patterns between the keywords related to some of the common disorders in neuroscience in an interactive way. PubMed database and keywords related to schizophrenia and alcoholism are used as inputs. This thesis reveals many associations between the different terms, which are otherwise difficult to understand by reading articles or journals manually. Graphviz tool is used to visualize associations.

Indiana University-Purdue University Indianapolis (IUPUI) / Joint models of longitudinal and survival outcomes have been used with increasing frequency in clinical investigations. Correct specification of fixed and random effects, as well as their functional forms is essential for practical data analysis. However, no existing methods have been developed to meet this need in a joint model setting. In this dissertation, I describe a penalized likelihood-based method with adaptive least absolute shrinkage and selection operator (ALASSO) penalty functions for model selection. By reparameterizing variance components through a Cholesky decomposition, I introduce a penalty function of group shrinkage; the penalized likelihood is approximated by Gaussian quadrature and optimized by an EM algorithm. The functional forms of the independent effects are determined through a procedure for structural discovery. Specifically, I first construct the model by penalized cubic B-spline and then decompose the B-spline to linear and nonlinear elements by spectral decomposition. The decomposition represents the model in a mixed-effects model format, and I then use the mixed-effects variable selection method to perform structural discovery. Simulation studies show excellent performance. A clinical application is described to illustrate the use of the proposed methods, and the analytical results demonstrate the usefulness of the methods.