This dissertation discusses three problems from different areas of medical research and their machine learning solutions. Each solution is a distinct type of decision support system. They show three common properties: personalized healthcare decision support, reduction of the use of medical resources, and improvement of outcomes. The first decision support system assists individual hospital selection. This system can help a user make the best decision in terms of the combination of mortality, complication, and travel distance. Both machine learning and optimization techniques are utilized in this type of decision support system. Machine learning methods, such as Support Vector Machines, learn a decision function. Next, the function is transformed into an objective function and then optimization methods are used to find the values of decision variables to reach the desired outcome with the most confidence. The second decision support system assists diagnostic decisions in a sequential decision-making setting by finding the most promising tests and suggesting a diagnosis. The system can speed up the diagnostic process, reduce overuse of medical tests, save costs, and improve the accuracy of diagnosis. In this study, the system finds the test most likely to confirm a diagnosis based on the pre-test probability computed from the patient's information including symptoms and the results of previous tests. If the patient's disease post-test probability is higher than the treatment threshold, a diagnostic decision will be made, and vice versa. Otherwise, the patient needs more tests to help make a decision. The system will then recommend the next optimal test and repeat the same process. The third decision support system recommends the best lifestyle changes for an individual to lower the risk of cardiovascular disease (CVD). As in the hospital recommendation system, machine learning and optimization are combined to capture the relationship between lifestyle and CVD, and then generate recommendations based on individual factors including preference and physical condition. The results demonstrate several recommendation strategies: a whole plan of lifestyle changes, a package of n lifestyle changes, and the compensatory plan (the plan that compensates for unwanted lifestyle changes or real-world limitations).

Drug effects can be unpredictable and vary widely among patients with environmental, genetic, and clinical factors. Randomized control trials (RCTs) are not sufficient to identify adverse drug reactions (ADRs), and the electronic health record (EHR) along with medical claims have become an important resource for pharmacovigilance. Among all the data collected in hospitals, laboratory tests represent the most documented and reliable data type in the EHR. Laboratory tests are at the core of the clinical decision process and are used for diagnosis, monitoring, screening, and research by physicians. They can be linked to drug effects either directly, with therapeutic drug monitoring (TDM), or indirectly using drug laboratory effects (DLEs) that affect surrogate tests. Unfortunately, very few automated methods use laboratory tests to inform clinical decision making and predict drug effects, partly due to the complexity of these time series that are irregularly sampled, highly dependent on other clinical covariates, and non-stationary. Deep learning, the branch of machine learning that relies on high-capacity artificial neural networks, has known a renewed popularity this past decade and has transformed fields such as computer vision and natural language processing. Deep learning holds the promise of better performances compared to established machine learning models, although with the necessity for larger training datasets due to their higher degrees of freedom. These models are more flexible with multi-modal inputs and can make sense of large amounts of features without extensive engineering. Both qualities make deep learning models ideal candidate for complex, multi-modal, noisy healthcare datasets. With the development of novel deep learning methods such as generative adversarial networks (GANs), there is an unprecedented opportunity to learn how to augment existing clinical dataset with realistic synthetic data and increase predictive performances. Moreover, GANs have the potential to simulate effects of individual covariates such as drug exposures by leveraging the properties of implicit generative models. In this dissertation, I present a body of work that aims at paving the way for next generation laboratory test-based clinical decision support systems powered by deep learning. To this end, I organized my experiments around three building blocks: (1) the evaluation of various deep learning architectures with laboratory test time series and their covariates with a forecasting task; (2) the development of implicit generative models of laboratory test time series using the Wasserstein GAN framework; (3) the inference properties of these models for the simulation of drug effects in laboratory test time series, and their application for data augmentation. Each component has its own evaluation: The forecasting task enabled me to explore the properties and performances of different learning architectures; the Wasserstein GAN models are evaluated with both intrinsic metrics and extrinsic tasks, and I always set baselines to avoid providing results in a "neural-network only" referential. Applied machine learning, and more so with deep learning, is an empirical science. While the datasets used in this dissertation are not publicly available due to patient privacy regulation, I described pre-processing steps, hyper-parameters selection and training processes with reproducibility and transparency in mind. In the specific context of these studies involving laboratory test time series and their clinical covariates, I found that for supervised tasks, machine learning holds up well against deep learning methods. Complex recurrent architectures like long short-term memory (LSTM) do not perform well on these short time series, while convolutional neural networks (CNNs) and multi-layer perceptrons (MLPs) provide the best performances, at the cost of extensive hyper-parameter tuning. Generative adversarial networks, enabled by deep learning models, were able to generate high-fidelity laboratory test time series, and the quality of the generated samples was increased with conditional models using drug exposures as auxiliary information. Interestingly, forecasting models trained on synthetic data exclusively still retain good performances, confirming the potential of GANs in privacy-oriented applications. Finally, conditional GANs demonstrated an ability to interpolate samples from drug exposure combinations not seen during training, opening the way for laboratory test simulation with larger auxiliary information spaces. In specific cases, augmenting real training sets with synthetic data improved performances in the forecasting tasks, and could be extended to other applications where rare cases present a high prediction error.

Abstract not available

Abstract not available

Abstract not available

A significant number of engineers move into management positions, their numbers increasing with their length of service. However, engineers are not considered to be effective managers and are generally considered inadequate in soft skills. Given the centrality of engineers and management, understanding this transition is essential in order to develop strategies for managing. This research is an exploratory field-based study of the transition of professional engineers into management roles (engineer-managers) in Australia, from the perspective of the individual engineer. The study investigates the attitudes of engineers towards such areas as engineering education, towards managerial transition, status, organizational support systems, and strategies for managing transition, and examines their influence on the process of transition. Importantly, this research examines the influence of factors such as job nature, management qualifications, age, employing organizations, and other variables on their attitudes, and studies the differences between various subgroups of engineers. This research is based on the results of a case study and a questionnaire survey. An important outcome of this research is the focus on the process of engineering education. This research concludes that different emphases in the process of teaching and learning would contribute, in the long run, to engineers developing soft skills, and so make their transition into management easier. The study found that electrical engineers are more proactive than civil or mechanical engineers and that it is necessary to develop different strategies for different groups of engineers. The study observed that the higher the status of professional engineers within an organization, the greater was the likelihood of success. Supporting the anecdotal evidence from the case study, it is noted that the more engineers there are in management positions, the better the perception of senior management about their capabilities. This study found that management education for engineers has a strong influence, both in terms of their acquiring managerial skills as well as enhancing their status within their organization. Experiential learning, though, is the most common method by which engineers acquire managerial skills. The study also found that this is the least-managed strategy in Australian organizations; learning is left entirely to the individual. For engineers to be able to take advantage of experiential learning, better management is necessary / Doctor of Philosophy (PhD)

This thesis has two distinct research threads. One thread examines the effectiveness of technology support on the performance of focus groups. Unlike previous research, the work described in this thesis addresses the fundamental issue that groups are social systems, and that comprehensive measurement of the effectiveness of group activities requires assessment of both the task-oriented and social aspects of the group activity. In this research, four different communication modes are used to compare group effectiveness. The second research thread in this thesis is the use of Systems Thinking, and specifically Soft Systems Methodology (SSM), as the framework for inquiring into the effects of technology support on group effectiveness. The strategy in this thesis for developing and evaluating hypotheses extends the general descriptions and guidance in the literature on using SSM for hypothesis testing. Systems thinking also provides the basis for examining the prevailing ???profile deviation??? view that the better the fit between the group task and the technology support the greater the group performance. Using the six perspectives of fit developed by Venkatraman (1989), the most common GSS models and other models developed to examine Task-Technology Fit (TTF) are analysed. The results show that group performance models are most often tested from a ???profile deviation??? perspective and TTF models developed from a profile deviation perspective claim to have predictive and descriptive validity for assessing the level of group performance. To assess whether an SSM based approach can improve the predictive and/or descriptive analysis of the impact of technology support on group work, a field experiment was conducted at the Australian Defence Force Academy. Twenty focus groups of officer cadets assessed their military training program using a GSS in one of four communication modes. The results showed little predictive or descriptive support for the profile deviation perspective of TTF when measuring the group???s overall effectiveness, task effectiveness, participant satisfaction or group relations. The alternative ???gestalt??? perspective, operationalised in this research by using SSM, provided a more comprehensive approach to examining the effectiveness of technology support for group work.

Existing patient medical records are a rich data source with a potential to support clinical research. Fragmentation of data across disparate medical database inhibits the use of these existing datasets. Overcoming such disjointedness is possible through the use of a data warehouse. Once the data is cleansed, transformed, and stored within the data warehouse it is possible to turn attention to the exploration of the medical datasets. Exploratory and confirmatory Data Mining Tools are well suited to such activities. This thesis concerned with: demonstrating parallels between scientific method and CRISP-DM; extending CRISP-DM for use with medical datasets; and proposal of the supporting Intelligent Decision Support System framework. This research has been undertaken using a fetal-maternal case study. / Master of Science (Hons)

Decision support systems (DSS) are usually based on computerised models of biophysical and economic systems. Despite early expectations that such models would inform and improve management, adoption rates have been low, and implementation of DSS is now “critical” The reasons for this are unclear and the aim of this study is to learn to better design, develop and apply DSS in farming systems research (FSR). Previous studies have explored the merits of quantitative tools including DSS, and suggested changes leading to greater impact. In Australia, the changes advocated have been: Simple, flexible, low cost economic tools: Emphasis on farmer learning through soft systems approaches: Understanding the socio-cultural contexts of using and developing DSS: Farmer and researcher co-learning from simulation modelling and Increasing user participation in DSS design and implementation. Twenty-four simple criteria were distilled from these studies, and their usefulness in guiding the development and application of DSS were assessed in six FSR case studies. The case studies were also used to better understand farmer learning through models of decision making and learning. To make DSS useful complements to farmers’ existing decision-making repertoires, they should be based on: (i) a decision-oriented development process, (ii) identifying a motivated and committed audience, (iii) a thorough understanding of the decision-makers context, (iv) using learning as the yardstick of success, and (v) understanding the contrasts, contradictions and conflicts between researcher and farmer decision cultures / Doctor of Philosophy (PhD)

Australian farmers are supplementing traditional practices with innovative strategies in an effort to survive recent economic, environmental, and social crises in the rural sector. These innovative strategies include moving towards a technology-based farm management style. A review of past literature determines that, despite a growing awareness of the usefulness of computers for farm management, there is concern over the limited demand for computer-based agricultural decision support systems (DSS). Recent literature indicates that women are the dominant users of computers on family farms yet are hesitant to use computers for decision support, and it is also unclear what decision-making roles women assume on family farms. While past research has investigated the roles of women in the Australian rural sector, there is a dearth of research into the interaction of women cotton growers with computers. Therefore, this dissertation is an ontological study and aims to contribute to scholarly knowledge in the research domain of Australian women cotton growers, agricultural DSS, and cotton farm management. This dissertation belongs in the Information Systems (IS) stream and describes an interpretive single case study which explores the lives of Australian women cotton growers on family farms and the association of an agricultural DSS with their farm management roles. Data collection was predominantly through semi-structured interviews with women cotton growers and cotton industry professionals such as DSS developers, rural extension officers, researchers and educators, rural experimental scientists, and agronomists and consultants, all of whom advise cotton growers. The study was informed by multiple sociological theories with opposing paradigmatic assumptions: Giddens' (1984) structuration theory as a metatheory to explore the recursiveness of farm life and technology usage; Rogers' (1995) diffusion of innovations theory with a functionalist approach to objectively examine the features of the software and user, as well as the processes of technology adoption; and Connell's (2002) theory of gender relations with its radical humanist perspective to subjectively investigate the relationships between farm partners through critical enquiry. The study was enriched further by drawing on other writings of these authors (Connell 1987; Giddens 2001; Rogers 2003) as well as complementary theories by authors (Orlikowski 1992; Orlikowski 2000; Trauth 2002; Vanclay & Lawrence 1995). These theories in combination have not been used before, which is a theoretical contribution of the study. The agricultural DSS for the study was CottonLOGIC, an advanced farm management tool to aid the management of cotton production. It was developed in the late 1990s by the CSIRO and the Australian Cotton Cooperative Research Centre (CRC), with support from the Cotton Research and Development Corporation (CRDC). CottonLOGIC is a software package of decision support and record-keeping modules to assist cotton growers and their advisors in the management of cotton pests, soil nutrition, and farm operations. It enables the recording and reporting of crop inputs and yields, insect populations (heliothis, tipworm, mirids and so on), weather data, and field operations such as fertiliser and pesticide applications, as well as the running of insect density prediction (heliothis and mites) and soil nutrition models. The study found that innovative practices and sustainable solutions are an imperative in cotton farm management for generating an improved triple bottom line of economic, environmental and social outcomes. CottonLOGIC is an industry benchmark for supporting these values through the incorporation of Best Management Practices (BMP) and Integrated Pest Management (IPM) principles, although there were indications that the software is in need of restructuring as could be expected of software over five years old. The evidence from the study was that women growers are participants in strategic farm decisions but less so in operational decisions, partly due to their lack of relevant agronomic knowledge. This hindered their use of CottonLOGIC, despite creative attempts to modify it. The study endorsed the existence of gender differences and inequalities in rural Australia. Nevertheless, the study also found that the women are valued for their roles as business partners in the multidisciplinary nature of farm management. All the same, there was evidence that greater collaboration and cooperation by farm partners and advisors would improve business outcomes. On the whole, however, women cotton growers are not passive agents but take responsibility for their own futures. In particular, DSS tools such as CottonLOGIC are instrumental in enabling women cotton growers to adapt to, challenge, and influence farm management practices in the family farm enterprise, just as CottonLOGIC is itself shaped and reshaped. Hence, a practical contribution of this study is to provide non-prescriptive guidelines for the improved adoption of agricultural DSS, particularly by rural women, as well as increasing awareness of the worth of their roles as family farm business partners.