For reasoning under uncertainty the Bayesian network has become the representation of choice. However, except where models are considered 'simple' the task of construction and inference are provably NP-hard. For modelling larger 'real' world problems this computational complexity has been addressed by methods that approximate the model. The Naive Bayes classifier, which has strong assumptions of independence among features, is a common approach, whilst the class of trees is another less extreme example. In this thesis we propose the use of an information theory based technique as a mechanism for inference in Singly Connected Networks. We call this a Mutual Information Measure classifier, as it corresponds to the restricted class of trees built from mutual information. We show that the new approach provides for both an efficient and localised method of classification, with performance accuracies comparable with the less restricted general Bayesian networks. To improve the performance of the classifier, we additionally investigate the possibility of expanding the class Markov blanket by use of a Wrapper approach and further show that the performance can be improved by focusing on the class Markov blanket and that the improvement is not at the expense of increased complexity. Finally, the two methods are applied to the task of diagnosing the 'real' world medical domain, Acute Abdominal Pain. Known to be both a different and challenging domain to classify, the objective was to investigate the optiniality claims, in respect of the Naive Bayes classifier, that some researchers have argued, for classifying in this domain. Despite some loss of representation capabilities we show that the Mutual Information Measure classifier can be effectively applied to the domain and also provides a recognisable qualitative structure without violating 'real' world assertions. In respect of its 'selective' variant we further show that the improvement achieves a comparable predictive accuracy to the Naive Bayes classifier and that the Naive Bayes classifier's 'overall' performance is largely due the contribution of the majority group Non-Specific Abdominal Pain, a group of exclusion.

Modeling species of conservation concern, such as those that are rare, declining, or have a conservation designation (e.g. endangered or threatened), remains an activity filled with uncertainty. Species that are of conservation concern often are found infrequently, in small sample sizes and spatially fragmented distributions, thereby making accurate enumeration difficult and traditional statistical approaches often invalid. For example, there are numerous debates in the ecological literature regarding methodological choices in conservation-based models, such as how to measure functional traits to account for ecosystem function, the impact of including rare species in biological assessments and whether species-specific dispersal can be measured using distance based functions. This thesis attempts to address issues in methodological choices in conservation-based models in two ways. In the first section of the thesis, the impacts of methodological choices on conservation-based models are examined across a broad selection of available approaches, from: measuring functional diversity; to conducting bio-assessments in community ecology; to assessing dispersal in metapopulation analyses. It is the goal of this section to establish the potential for methodological choices to impact conservation-based models, regardless of the scale, study-system or species involved. In the second section of this thesis, the use of consensus methods is developed as a potential tool for reducing uncertainty with methodological choices in conservation-based models. Two separate applications of consensus methods are highlighted, including how consensus methods can reduce uncertainty from choosing a modeling type or to identify when methodological choices may be a problem.

No description available.

Injury now surpasses disease as the leading global cause of premature death and disability, claiming over 5.8 millions lives each year. However, unlike disease, which has been subjected to a rigorous epidemiologic approach, the field of injury prevention and control has been a relative newcomer to scientific investigation. With the distribution of injury now well described (i.e. ???who???, ???what???, ???where??? and ???when???), the underlying hypothesis is that progress in understanding ???how??? and ???why??? lies in classifying injury occurrences aetiologically. The advancement of a means of classifying injury aetiology has so far been inhibited by two related limitations: 1. Structural limitation: The absence of a cohesive and validated aetiological taxonomy for injury, and; 2. Methodological limitation: The need to manually classify large numbers of injury cases to determine aetiological patterns. This work is directed at overcoming these impediments to injury research. An aetiological taxonomy for injury was developed consistent with epidemiologic principles, along with clear conventions and a defined three-tier hierarchical structure. Validation testing revealed that the taxonomy could be applied with a high degree of accuracy (coder/gold standard agreement was 92.5-95.0%), and with high inter- and intra- coder reliability (93.0-96.3% and 93.5-96.3%). Practical application demonstrated the emergence of strong aetiological patterns which provided insight into causative sequences leading to injury, and led to the identification of effective control measures to reduce injury frequency and severity. However, limitations related to the inefficient and error-prone manual classification process (i.e. average 4.75 minute/case processing time and 5.0-7.5% error rate), revealed the need for an automated approach. To overcome these limitations, a knowledge acquisition (KA) software tool was developed, tested and applied, based on an expertsystems technique known as ripple down rules (RDR). It was found that the KA system was able acquire tacit knowledge from a human expert and apply learned rules to efficiently and accurately classify large numbers of injury cases. Ultimately, coding error rates dropped to 3.1%, which, along with an average 2.50 minute processing time, compared favourably with results from manual classification. As such, the developed taxonomy and KA tool offer significant advantages to injury researchers who have a need to deduce useful patterns from injury data and test hypotheses regarding causation and prevention.

Diss. (sammanfattning) Umeå : Sveriges lantbruksuniv., 2007. / Härtill 5 uppsatser.

Diss. (sammanfattning) Umeå : Sveriges lantbruksuniversitet, 2006. / Härtill 4 uppsatser.

Licentiatavhandling (sammanfattning) Uppsala : Sveriges lantbruksuniverstet, 2006. / Härtill 2 uppsatser.

Diss. (sammanfattning) Umeå : Sveriges lantbruksuniv., 2009. / Härtill 3 uppsatser.

Diss. (sammanfattning) Umeå : Sveriges lantbruksuniv., 2009. / Härtill 6 uppsatser.

Diss. (sammanfattning) Uppsala : Sveriges lantbruksuniversitet, 2008. / Härtill 4 uppsatser.