Scheduling Optimization with LDA and Greedy Algorithm

Scheduling optimization is concerned with the optimal allocation of events to time slots. In this paper, we look at one particular example of scheduling problems - the 2015 Joint Statistical Meetings. We want to assign each session among similar topics to time slots to reduce scheduling conflicts. Chapter 1 briefly talks about the motivation for this example as well as the constraints and the optimality criterion. Chapter 2 proposes use of Latent Dirichlet Allocation (LDA) to identify the topic proportions in each session and talks about the fitting of the model. Chapter 3 translates these ideas into a mathematical formulation and introduces a Greedy Algorithm to minimize conflicts. Chapter 4 demonstrates the improvement of the scheduling with this method.

Optimized approach to decision fusion of heterogeneous data for breast cancer diagnosis.

As more diagnostic testing options become available to physicians, it becomes more difficult to combine various types of medical information together in order to optimize the overall diagnosis. To improve diagnostic performance, here we introduce an approach to optimize a decision-fusion technique to combine heterogeneous information, such as from different modalities, feature categories, or institutions. For classifier comparison we used two performance metrics: The receiving operator characteristic (ROC) area under the curve [area under the ROC curve (AUC)] and the normalized partial area under the curve (pAUC). This study used four classifiers: Linear discriminant analysis (LDA), artificial neural network (ANN), and two variants of our decision-fusion technique, AUC-optimized (DF-A) and pAUC-optimized (DF-P) decision fusion. We applied each of these classifiers with 100-fold cross-validation to two heterogeneous breast cancer data sets: One of mass lesion features and a much more challenging one of microcalcification lesion features. For the calcification data set, DF-A outperformed the other classifiers in terms of AUC (p < 0.02) and achieved AUC=0.85 +/- 0.01. The DF-P surpassed the other classifiers in terms of pAUC (p < 0.01) and reached pAUC=0.38 +/- 0.02. For the mass data set, DF-A outperformed both the ANN and the LDA (p < 0.04) and achieved AUC=0.94 +/- 0.01. Although for this data set there were no statistically significant differences among the classifiers' pAUC values (pAUC=0.57 +/- 0.07 to 0.67 +/- 0.05, p > 0.10), the DF-P did significantly improve specificity versus the LDA at both 98% and 100% sensitivity (p < 0.04). In conclusion, decision fusion directly optimized clinically significant performance measures, such as AUC and pAUC, and sometimes outperformed two well-known machine-learning techniques when applied to two different breast cancer data sets.

The environmental justice dimensions of climate change

Nations around the world are considering strategies to mitigate the severe impacts of climate change predicted to occur in the twenty-first century. Many countries, however, lack the wealth, technology, and government institutions to effectively cope with climate change. This study investigates the varying degrees to which developing and developed nations will be exposed to changes in three key variables: temperature, precipitation, and runoff. We use Geographic Information Systems (GIS) analysis to compare current and future climate model predictions on a country level. We then compare our calculations of climate change exposure for each nation to several metrics of political and economic well-being. Our results indicate that the impacts of changes in precipitation and runoff are distributed relatively equally between developed and developing nations. In contrast, we confirm research suggesting that developing nations will be affected far more severely by changes in temperature than developed nations. Our results also suggest that this unequal impact will persist throughout the twenty-first century. Our analysis further indicates that the most significant temperature changes will occur in politically unstable countries, creating an additional motivation for developed countries to actively engage with developing nations on climate mitigation strategies. © 2011, Mary Ann Liebert, Inc.

Using decision analysis to improve malaria control policy making.

Malaria and other vector-borne diseases represent a significant and growing burden in many tropical countries. Successfully addressing these threats will require policies that expand access to and use of existing control methods, such as insecticide-treated bed nets (ITNs) and artemesinin combination therapies (ACTs) for malaria, while weighing the costs and benefits of alternative approaches over time. This paper argues that decision analysis provides a valuable framework for formulating such policies and combating the emergence and re-emergence of malaria and other diseases. We outline five challenges that policy makers and practitioners face in the struggle against malaria, and demonstrate how decision analysis can help to address and overcome these challenges. A prototype decision analysis framework for malaria control in Tanzania is presented, highlighting the key components that a decision support tool should include. Developing and applying such a framework can promote stronger and more effective linkages between research and policy, ultimately helping to reduce the burden of malaria and other vector-borne diseases.

Factors influencing malaria control policy-making in Kenya, Uganda and Tanzania.

BACKGROUND: Policy decisions for malaria control are often difficult to make as decision-makers have to carefully consider an array of options and respond to the needs of a large number of stakeholders. This study assessed the factors and specific objectives that influence malaria control policy decisions, as a crucial first step towards developing an inclusive malaria decision analysis support tool (MDAST). METHODS: Country-specific stakeholder engagement activities using structured questionnaires were carried out in Kenya, Uganda and Tanzania. The survey respondents were drawn from a non-random purposeful sample of stakeholders, targeting individuals in ministries and non-governmental organizations whose policy decisions and actions are likely to have an impact on the status of malaria. Summary statistics across the three countries are presented in aggregate. RESULTS: Important findings aggregated across countries included a belief that donor preferences and agendas were exerting too much influence on malaria policies in the countries. Respondents on average also thought that some relevant objectives such as engaging members of parliament by the agency responsible for malaria control in a particular country were not being given enough consideration in malaria decision-making. Factors found to influence decisions regarding specific malaria control strategies included donor agendas, costs, effectiveness of interventions, health and environmental impacts, compliance and/acceptance, financial sustainability, and vector resistance to insecticides. CONCLUSION: Malaria control decision-makers in Kenya, Uganda and Tanzania take into account health and environmental impacts as well as cost implications of different intervention strategies. Further engagement of government legislators and other policy makers is needed in order to increase funding from domestic sources, reduce donor dependence, sustain interventions and consolidate current gains in malaria.

VizMaps: A Bayesian Topic Modeling Based PubMed Search Interface

A common challenge that users of academic databases face is making sense of their query outputs for knowledge discovery. This is exacerbated by the size and growth of modern databases. PubMed, a central index of biomedical literature, contains over 25 million citations, and can output search results containing hundreds of thousands of citations. Under these conditions, efficient knowledge discovery requires a different data structure than a chronological list of articles. It requires a method of conveying what the important ideas are, where they are located, and how they are connected; a method of allowing users to see the underlying topical structure of their search. This paper presents VizMaps, a PubMed search interface that addresses some of these problems. Given search terms, our main backend pipeline extracts relevant words from the title and abstract, and clusters them into discovered topics using Bayesian topic models, in particular the Latent Dirichlet Allocation (LDA). It then outputs a visual, navigable map of the query results.