Melbourne School of Population and Global Health - Research Publications
Permanent URI for this collection
Search Results
1 - 10 of 41
-
ItemAn Uncertainty-Accuracy-Based Score Function for Wrapper Methods in Feature Selection(IEEE, 2023-08-13)Feature Selection (FS) is an effective preprocessing method to deal with the curse of dimensionality in machine learning. Redundant features in datasets decrease the classification performance and increase the computational complexity. Wrapper methods are an important category of FS methods that evaluate various feature subsets and select the best one using performance measures related to a classifier. In these methods, the accuracy of classifiers is the most common performance measure for FS. Although the performance of classifiers depends on their uncertainty, this important criterion is neglected in these methods. In this paper, we present a new performance measure called Uncertainty-Accuracy-based Performance Measure for Feature Selection (UAPMFS) that uses an ensemble approach to measure both the accuracy and uncertainty of classifiers. UAPMFS uses bagging and uncertainty confusion matrix. This performance measure can be used in all wrapper methods to improve FS performance. We design two experiments to evaluate the performance of UAPMFS in wrapper methods. In experiments, we use the leave-one-variable-out strategy as the common strategy in wrapper methods to evaluate features. We also define a feature score function based on UAPMFS to rank and select features. In the first experiment, we investigate the importance of considering uncertainty in the FS process and show how neglecting uncertainty affects FS performance. In the second experiment, we compare the performance of the UAPMFS-based feature score function with the most common feature score functions for FS. Experimental results show the effectiveness of the proposed performance measure on different datasets.
-
ItemCluster-based Diversity Over-sampling: A Density and Diversity Oriented Synthetic Over-sampling for Imbalanced Data(SCITEPRESS - Science and Technology Publications, 2022)In many real-life classification tasks, the issue of imbalanced data is commonly observed. The workings of mainstream machine learning algorithms typically assume the classes amongst underlying datasets are relatively well-balanced. The failure of this assumption can lead to a biased representation of the models’ performance. This has encouraged the incorporation of re-sampling techniques to generate more balanced datasets. However, mainstream re-sampling methods fail to account for the distribution of minority data and the diversity within generated instances. Therefore, in this paper, we propose a data-generation algorithm, Cluster-based Diversity Over-sampling (CDO), to consider minority instance distribution during the process of data generation. Diversity optimisation is utilised to promote diversity within the generated data. We have conducted extensive experiments on synthetic and real-world datasets to evaluate the performance of CDO in comparison with SMOTE-based and diversity-based methods (DADO, DIWO, BL-SMOTE, DB-SMOTE, and MAHAKIL). The experiments show the superiority of CDO.
-
ItemReal-time Repatriation: Data Governance for Social Anthropology in the 21st Century(Paris Nanterre University, 2022)Social Anthropologists are currently grappling with complex simultaneous changes in research ethics and data governance regimes across diverse jurisdictions. Internationally, repatriation of unethically acquired ethnographic collections is becoming common-place, resulting in the return of both tangible and intangible cultural assets to their rightful owners. In Europe, the open data movement and the recent implementation of the General Data Protection Regulation appears to challenge social anthropologists’ commitment to protect the confidentiality of oftenvulnerable research participants. Meanwhile, in Australia, New Zealand, Canada and elsewhere, the popular refrain ‘Indigenous data sovereignty’ is compelling many social anthropologists to face discomforting aspects of their field’s involvement in colonial administrative regimes. In this paper I present a model for how social anthropology might reconcile three interrelated factors contributing to this complex situation: 1) Repatriation of data collected from research participants without clear or sufficiently comprehensive consent; 2) Risks and opportunities presented by legislated instances of the open data movement and; 3) Relevant and operable features of the Indigenous data sovereignty movement. Drawing on my 20-year career as a forensic and expert social anthropologist working with Indigenous community organisations on land rights and cultural heritage preservation cases in the Australian Federal Court and under Northern Territory statutory regimes, I illustrate how this model of social anthropological data governance can be put into effect. ‘Real-time repatriation’ describes the synthesis of leading ethical, legal and technological standards in proactively upholding and safeguarding the interests and decision-making autonomy of participants in social anthropological research.
-
ItemForensic and Expert Social Anthropology as Cultural Expertise: Developing Professional Standards(Paris-Sorbonne University, 2022)Social anthropology has served as a specific form of cultural expertise in legal-administrative processes since the late 19th century. Since the 1970s, social anthropology has served increasingly as a source of forensic and expert evidence in legal processes in particular, including asylum claims, traditional land ownership and cultural heritage restitution claims, human rights abuse inquiries, and international development initiatives, among others. Notwithstanding this increasing specialisation, internationally recognised professional standards for the provision of forensic and expert social anthropological evidence have remained historically absent. In response, in late 2021, the Royal Anthropological Institute (RAI) commenced development of a professional standards and certification framework for social anthropologists engaged in the provision of forensic and expert evidence to courts and other legally empowered bodies. The RAI initiative represents the first internationally coordinated effort by a professional social anthropological representative organisation to develop such standards. The RAI framework provides a mechanism by which social anthropologists may engage in the provision of forensic and expert services in a relevant, systematic and professionally recognised manner across a range of legal-administrative jurisdictions. The RAI framework establishes terms and definitions for use by both social anthropologists and legal professionals in assessing the prospective relevance of social anthropological services to legal-administrative processes, and in deciding the form that such services should take. This paper will provide a summary of the RAI forensic and expert social anthropology framework and its particular relevance to legal-administrative processes as a specific form of cultural expertise more broadly.
-
Item
-
ItemSteroid hormone measurements from different types of assays in relation to body mass index and breast cancer risk in postmenopausal women: Reanalysis of eighteen prospective studies(ELSEVIER SCIENCE INC, 2015-07)Epidemiological studies have examined breast cancer risk in relation to sex hormone concentrations measured by different methods: "extraction" immunoassays (with prior purification by organic solvent extraction, with or without column chromatography), "direct" immunoassays (no prior extraction or column chromatography), and more recently with mass spectrometry-based assays. We describe the associations of estradiol, estrone and testosterone with both body mass index and breast cancer risk in postmenopausal women according to assay method, using data from a collaborative pooled analysis of 18 prospective studies. In general, hormone concentrations were highest in studies that used direct assays and lowest in studies that used mass spectrometry-based assays. Estradiol and estrone were strongly positively associated with body mass index, regardless of the assay method; testosterone was positively associated with body mass index for direct assays, but less clearly for extraction assays, and there were few data for mass spectrometry assays. The correlations of estradiol with body mass index, estrone and testosterone were lower for direct assays than for extraction and mass spectrometry assays, suggesting that the estimates from the direct assays were less precise. For breast cancer risk, all three hormones were strongly positively associated with risk regardless of assay method (except for testosterone by mass spectrometry where there were few data), with no statistically significant differences in the trends, but differences may emerge as new data accumulate. Future epidemiological and clinical research studies should continue to use the most accurate assays that are feasible within the design characteristics of each study.
-
ItemNo Preview AvailableData Governance and Social Anthropology(European Association of Social Anthropologists, 2021-12-09)
-
ItemCollaborative Human-ML Decision Making Using Experts' Privileged Information under Uncertainty( 2021-01-01)Machine Learning (ML) models have been widely applied for clinical decision making. However, in this critical decision making field, human decision making is still prevalent, because clinical experts are more skilled to work with unstructured data specially to deal with uncommon situations. In this paper, we use clinical experts' privileged information as an information source for clinical decision making besides information provided by ML models and introduce a collaborative human-ML decision making model. In the proposed model, two groups of decision makers including ML models and clinical experts collaborate to make a consensus decision. As decision making always comes with uncertainty, we present an interval modelling to capture uncertainty in the proposed collaborative model. For this purpose, clinical experts are asked to give their opinion as intervals, and we generate prediction intervals as the outputs of ML models. Using Interval Agreement Approach (IAA), as an aggregation function in our proposed collaborative model, pave the way to minimize loss of information through aggregating intervals to a fuzzy set. The proposed model not only can improve the accuracy and reliability of decision making, but also can be more interpretable especially when it comes to critical decisions. Experimental results on synthetic data shows the power of the proposed collaborative decision making model in some scenarios.
-
ItemIncreases in controlled-release oxycodone utilisation following the subsidy of oxycodone with naloxone formulations: An Australian population-based study(WILEY, 2019-01)PURPOSE: Despite increasing use of oxycodone/naloxone controlled-release (CR) in Australia, little is known about how it has affected the overall oxycodone CR market since its subsidy in 2011. METHODS: We used Pharmaceutical Benefits Scheme dispensing claims (2006-2016) and interrupted time series analysis to examine changes in the quarterly rates of dispensing of oral oxycodone CR formulations (oxycodone/naloxone CR and single-ingredient oxycodone CR) and new oxycodone CR treatment episodes. We also performed a retrospective cohort study in a sample of people initiating a new oxycodone CR treatment episode in 2009, 2012/2013, and 2016 to compare opioid utilisation patterns over time. RESULTS: The subsidy of oxycodone/naloxone CR was associated with a 1.6-fold increase in the growth rate of oxycodone CR dispensing, resulting from rapid uptake of low strength (≤5 mg) oxycodone/naloxone CR. In our cohort of initiators, the number of new oxycodone CR treatment episodes increased 2.1-fold between 2009 and 2016; in 2016, 91.4% of new treatment episodes involved oxycodone/naloxone CR. Comparing 2016 with 2009, we observed an increase in people initiating with a tablet strength less than or equal to 5-mg (risk difference [RD] = 21.1%, 95% CI, 19.9%-22.4%) in people initiating with no other opioid dispensing 90 days prior to initiation (RD = 5.2%, 3.8%-6.6%) and with no further opioid dispensing 90 days after initiation (RD = 8.8%, 7.4%-10.2%). CONCLUSIONS: After its subsidy, the uptake of low-dose oxycodone/naloxone CR was greater than expected if it were substituting the single-ingredient oxycodone CR, resulting in an expansion of the oxycodone CR market.
-
ItemInjecting drug use is an independent risk factor for reincarceration after release from prison: A prospective cohort study(WILEY, 2019-03)INTRODUCTION AND AIMS: Once involved in the criminal justice system, people who inject drugs (PWID) have a high probability of multiple system encounters. Imprisonment typically fails to rehabilitate PWID, who upon return to the community are at considerable risk of returning to injecting drug use (IDU) and poor health and social outcomes. We examined the effect of IDU resumption, and a suite of other sociodemographic, criminogenic, health and behavioural indicators, on the timing of reincarceration among adults with a history of IDU following release from prison. DESIGN AND METHODS: Structured interviews were conducted with 561 PWID in Queensland, Australia prior to release from prison and approximately 1, 3 and 6 months post-release. Data were linked prospectively with correctional records and the National Death Index. Data collected at multiple time-points were treated as time-varying covariates. Kaplan-Meier survival estimates and Cox proportional hazards models were used to estimate the rate and hazards of reincarceration. RESULTS: Sixty-eight percent of participants (n = 350) were reincarcerated over a combined observation time of 1043.5 years, representing a rate of 33.5 per 100 person-years (95% confidence interval [CI] 30.2-37.2). Time-invariant predictors of reincarceration in PWID were: male gender (adjusted hazard ratio [AHR] = 1.62, 95% CI 1.19-2.21), older age at release (AHR = 0.97, 95% CI 0.95-1.00), previous adult (AHR = 2.00, 95% CI 1.41-2.84) or juvenile (AHR = 1.78, 95% CI 1.27-2.49) imprisonment, shorter imprisonment (≤90 days vs. >365 days, AHR = 2.09, 95% CI 1.30-3.34), release on parole (AHR = 2.29, 95% CI 1.82-2.88) and drug-related sentence (AHR = 1.84, 95% CI 1.34-2.53). Time-varying predictors included resumption of IDU (AHR = 2.04, 95% CI 1.60-2.61), unemployment (AHR = 1.53, 95% CI 1.07-2.19) and low perceived social support (AHR = 1.41, 95% CI 1.05-1.90). Very-high psychological distress at the most recent interview was protective against reincarceration (AHR = 0.65, 95% CI 0.44-0.95). DISCUSSION AND CONCLUSIONS: Efforts to prevent resumption of IDU and address disadvantage, social inclusion and health service access in ex-prisoners through the scale-up and integration of prison-based and post-release interventions are likely to reap both public health and criminal justice benefits.
- 1 (current)
- 2
- 3
- 4
- 5