Evaluation of sampling methods for learning from imbalanced data

The problem of learning from imbalanced data is of critical importance in a large number of application domains and can be a bottleneck in the performance of various conventional learning methods that assume the data distribution to be balanced. The class imbalance problem corresponds to dealing with the situation where one class massively outnumbers the other. The imbalance between majority and minority would lead machine learning to be biased and produce unreliable outcomes if the imbalanced data is used directly. There has been increasing interest in this research area and a number of algorithms have been developed. However, independent evaluation of the algorithms is limited. This paper aims at evaluating the performance of five representative data sampling methods namely SMOTE, ADASYN, BorderlineSMOTE, SMOTETomek and RUSBoost that deal with class imbalance problems. A comparative study is conducted and the performance of each method is critically analysed in terms of assessment metrics. © 2013 Springer-Verlag.

The Management of Long-Term Sickness Absence in Large Public Sector Healthcare Organisations: A Realist Evaluation Using Mixed Methods

Purpose The success of measures to reduce long-term sickness absence (LTSA) in public sector organisations is contingent on organisational context. This realist evaluation investigates how interventions interact with context to influence successful management of LTSA. Methods Multi-method case study in three Health and Social Care Trusts in Northern Ireland comprising realist literature review, semi-structured interviews (61 participants), Process-Mapping and feedback meetings (59 participants), observation of training, analysis of documents. Results Important activities included early intervention; workplace-based occupational rehabilitation; robust sickness absence policies with clear trigger points for action. Used appropriately, in a context of good interpersonal and interdepartmental communication and shared goals, these are able to increase the motivation of staff to return to work. Line managers are encouraged to take a proactive approach when senior managers provide support and accountability. Hindering factors: delayed intervention; inconsistent implementation of policy and procedure; lack of resources; organisational complexity; stakeholders misunderstanding each other’s goals and motives. Conclusions Different mechanisms have the potential to encourage common motivations for earlier return from LTSA, such as employees feeling that they have the support of their line manager to return to work and having the confidence to do so. Line managers’ proactively engage when they have confidence in the support of seniors and in their own ability to address LTSA. Fostering these motivations calls for a thoughtful, diagnostic process, taking into account the contextual factors (and whether they can be modified) and considering how a given intervention can be used to trigger the appropriate mechanisms.

An Evaluation of 1D Design Methods for the Off-Design Performance Prediction of Automotive Turbocharger Compressors

Several one-dimensional design methods have been used to predict the off-design performance of three modern centrifugal compressors for automotive turbocharging. The three methods used are single-zone, two-zone, and a more recent statistical method. The predicted results from each method are compared against empirical data taken from standard hot gas stand tests for each turbocharger. Each of the automotive turbochargers considered in this study have notably different geometries and are of varying application. Due to the non-adiabatic test conditions, the empirical data has been corrected for the effect of heat transfer to ensure comparability with the 1D models. Each method is evaluated for usability and accuracy in both pressure ratio and efficiency prediction. The paper presents an insight into the limitations of each of these models when applied to one-dimensional automotive turbocharger design, and proposes that a corrected single-zone modelling approach has the greatest potential for further development, whilst the statistical method could be immediately introduced to a design process where design variations are limited.

An Evaluation of Vaneless Diffuser Modelling Methods as a Means of Improving the Off-Design Performance Prediction of Centrifugal Compressors

An evaluation of existing 1-D vaneless diffuser design tools in the context of improving the off-design performance prediction of automotive turbocharger centrifugal compressors is described. A combination of extensive gas stand test data and single passage CFD simulations have been employed in order to permit evaluation of the different methods, allowing conclusions about the relative benefits and deficiencies of each of the different approaches to be determined. The vaneless diffuser itself has been isolated from the incumbent limitations in the accuracy of 1-D impeller modelling tools through development of a method to fully specify impeller exit conditions (in terms of mean quantities) using only standard test stand data with additional interstage static pressure measurements at the entrance to the diffuser. This method allowed a direct comparison between the test data and 1-D methods through sharing common inputs, thus achieving the aim of diffuser isolation.

Crucial to the accuracy of determining the performance of each of the vaneless diffuser configurations was the ability to quantify the presence and extent of the spanwise aerodynamic blockage present at the diffuser inlet section. A method to evaluate this critical parameter using CFD data is described herein, along with a correlation for blockage related to a new diffuser inlet flow parameter ⚡, equal to the quotient of the local flow coefficient and impeller tip speed Mach number. The resulting correlation permitted the variation of blockage with operating condition to be captured.

An Evaluation of the Impact of the Prospective Payment System on Antidepressant Use in Nursing Home Residents

Objectives: To determine the impact of the prospective payment system (PPS) for skilled nursing facilities on the pharmacologic treatment of depression.

Methods: We used a quasi-experimental study comparing the pharmacological treatment rates for depression in the pre-PPS period (1997) to the post-PPS period (2000) in 8149 residents with documented depression living in over 500 nursing facilities in Ohio. Logistic regression models adjusting for clustering effects of residents residing in homes using generalized estimating equations provided estimates of the PPS effect on use of any antidepressant and the use of selective serotonin reuptake inhibitors (SSRIs). We evaluated the extent to which the PPS effect was modified by organizational characteristics, including structural characteristics, resource characteristics, and staff resources available in the homes.

Results: Overall, there was no difference in the likelihood of any antidepressant [odds ratio (OR), 1.05; 95% confidence interval (CI), 0.93 to 1.18, resident-adjusted model] or an SSRI being used (OR, 0.98; 95% CI, 0.86 to 1.12, resident-adjusted model) after the introduction of PPS compared with 1997 when this reimbursement system was not in place (referent group). These trends did not appear to be modified substantially by organizational characteristics.

Conclusion: Although PPS did not appear to have influenced the treatment of depression in nursing homes, systems that provide checks and balances in relation to PPS are warranted.

Evaluation of a Water-soluble Bioadhesive Patch for Photodynamic Therapy of Vulval Lesions

An innovative bioadhesive patch intended primarily as a vulval drug delivery system and, specifically, as a means to deliver photosensitisers, or their prodrugs, for photodynamic purposes is described. The patch was formulated with a copolymer of methyl vinyl ether and maleic anhydride (PMVE/MA) as a bioadhesive matrix and poly(vinyl chloride) as a drug-impervious backing layer. Adhesive strength to neonate porcine skin, as a model substrate, was evaluated using peel and tensile testing measurements. Acceptabilities of non-drug loaded patches were appraised using human volunteers and visual-analogue scoring devices. An optimal formulation, with water uptake and peel strengths appropriate for vulval drug delivery, was cast from a 20% (w/w) PMVE/MA solution and adhered with a strength of approximately 1.7 N cm-2. Patient evaluation demonstrated comfort and firm attachment for up to 4 h in mobile patients. Aminolevulinic acid, a commonly used photosensitiser, was formulated into the candidate formulation and applied to vulval intraepithelial neoplastic lesions. Fluorescence under ultraviolet illumination revealed protoporphyrin synthesis. The patch achieves the extended application times obligatory in topical photodynamic therapy of vulval lesions, thereby contributing to potential methods for the eradication of neoplastic lesions in the lower female reproductive tract.

Predictors of therapy resistant asthma: outcome of a systematic evaluation protocol

Background: It has been suggested that asthmatic subjects with persisting symptoms despite adequate maintenance therapy should be systematically evaluated to identify factors contributing to poor control. The aims of this study were to examine the prevalence of these factors in a cohort of sequentially referred poorly controlled asthmatics, and to determine if any factor or combination of factors predicted true therapy resistant asthma (TRA).

Methods: Patients were evaluated using a systematic evaluation protocol including induced sputum analysis, psychiatric assessment, ear, nose and throat examination, pulmonary function testing, high resolution CT scan of the thorax, and 24 hour dual probe ambulatory oesophageal pH monitoring; any identified provoking factor was treated. Asthma was managed according to BTS guidelines.

Results: Of 73 subjects who completed the assessment, 39 responded to intervention and 34 had TRA. Subjects with TRA had a greater period of instability, a higher dose of inhaled steroids at referral, more rescue steroid use, and a lower best percentage forced expiratory volume in 1 second (FEV1%). Oesophageal reflux, upper airway disease, and psychiatric morbidity were common (57%, 95%, 49%, respectively) but were not more prevalent in either group. Using multivariate logistic regression analysis, inhaled steroid dose >2000 µg BDP, previous assessment by a respiratory specialist, and initial FEV1% of <70% at referral predicted a final diagnosis of TRA.

Conclusions: In poorly controlled asthmatics there is a high prevalence of co-morbidity, identified by detailed systematic assessment, but no difference in prevalence between those who respond to intervention and those with TRA. Targeted treatment of identified co-morbidities has minimal impact on asthma related quality of life in those with therapy resistant disease.

Sex Education in Northern Ireland Schools: a Critical Evaluation.

To date there has been little research on young people and sexuality in Northern Ireland. This paper draws on the first major study in this area to analyse the delivery of formal sex education in schools. Both quantitative and qualitative methods were used to access young people's opinions about the quality of the sex education they had received at school. Overall, they reported high levels of dissatisfaction, with notable variations in relation to both gender and religious affiliation. In one sense their opinions mesh well with those of young people in other parts of these islands. At the same time the specificity of sexuality in Ireland plays a key role in producing the moral system that underlies much of formal sex education in schools. Underpinned by a particularly traditional and conservative strain of Christian morality, sex education in Northern Ireland schools is marked by conservatism and silence and by the avoidance of opportunities for informed choice in relation to sexuality on the part of young people.

Evaluation of predicted network modules in yeast metabolism using NMR-based metabolite profiling

Genome-scale metabolic models promise important insights into cell function. However, the definition of pathways and functional network modules within these models, and in the biochemical literature in general, is often based on intuitive reasoning. Although mathematical methods have been proposed to identify modules, which are defined as groups of reactions with correlated fluxes, there is a need for experimental verification. We show here that multivariate statistical analysis of the NMR-derived intra- and extracellular metabolite profiles of single-gene deletion mutants in specific metabolic pathways in the yeast Saccharomyces cerevisiae identified outliers whose profiles were markedly different from those of the other mutants in their respective pathways. Application of flux coupling analysis to a metabolic model of this yeast showed that the deleted gene in an outlying mutant encoded an enzyme that was not part of the same functional network module as the other enzymes in the pathway. We suggest that metabolomic methods such as this, which do not require any knowledge of how a gene deletion might perturb the metabolic network, provide an empirical method for validating and ultimately refining the predicted network structure.

An evaluation of cyclooxygenase-2 as a prognostic biomarker in mid-gut carcinoid tumours

Background/Aims:Mid-gut carcinoids (MGC) are the most common of the gastrointestinal carcinoid tumours. There is a lack of reliable prognostic indicators for MGC. Cox-2 and Bcl-2 were evaluated as prognostic biomarkers in a cohort of well-characterised non-appendiceal MGC. Methods: Tissue from the primary MGC tumours of 37 patients was subjected to immunohistochemical detection of Cox-2 and Bcl-2. In 9 cases tissue from secondary lesions was also examined. The study assessed whether tumour-associated Cox-2 and Bcl-2 expression were related to patient survival. Results: Cox-2 expression was demonstrated in 30/36 primary tumours. When all tumours were analysed Cox-regression analysis indicated a trend towards worsening survival with increasing Cox-2 histoscore (intensity x proportion; hazard ratio 1.53, 95%CI 0.93, 2.52; p=0.09). Analysis of Cox-2 positive tumours revealed a highly significant association between increasing histoscore and decreased survival (hazard ratio 3.03, 95%CI 1.33, 6.91, p=0.008). Tumour-associated Bcl-2 expression had no effect on patient survival (hazard ratio 1.12, 95% CI 0.42, 2.99 p=0.82). There was no significant association between Cox-2 and Bcl-2 expression (ï?£2 p=0.16), or Cox-2 histoscore and Bcl-2 expression (MWU p=0.59). Analysis of the Cox-2 histoscores of primary tumours and their corresponding secondary lesions, revealed a statistically significant trend towards increasing histoscore in the latter (Wilcoxon p=0.04). Conclusions: This study has provided evidence that Cox-2 expression in primary MGC may be associated with a more negative prognostic outlook.

Comparative evaluation of 2,3-bis[2-methyloxy-4-nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilide (XTT) and 2-(2-methoxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2, 4-disulfophenyl)2H-tetrazolium, monosodium salt (WST-8) rapid colorimetric assays for antimicrobial susceptibility testing of staphylococci and ESBL-producing clinical isolates

A new generation of water soluble tetrazolium salts have recently become available and in this study we compared a colorimetric assay developed using one of these salts, 2-(2-methoxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2, 4-disulfophenyl)-2H-tetrazolium, monosodium salt (WST-8), with a previously developed 2,3-bis[2-methyloxy-4-nitro-5-sulfophenyl]-2H-tetrazolium-5-carboxanilide(XTT) colorimetric assay to determine which agent is most suitable for use as a colorimetric indicator in susceptibility testing. The MICs of 6 antibiotics were determined for 33 staphylococci using both colorimetric assays and compared with those obtained using the British Society for Antimicrobial Chemotherapy reference broth microdilution method. Absolute categorical agreement between the reference and test methods ranged from 79% (cefuroxime) to 100% (vancomycin) for both assays. No minor or major errors occurred using either assay with very major errors ranging from zero (vancomycin) to seven (cefuroxime). Analysis of the distribution of differences in the 1092 dilution MIC results revealed overall agreement, within the accuracy limits of the standard test ( 1 1092 dilution), using the XTT and WST-8 assays of 98% and 88%, respectively. Further studies on 31 ESBL-producing isolates were performed using the XTT method with absolute categorical agreement ranging from 87% (nitrofurantoin) to 100% (ofloxacin and meropenem). No errors were noted for either ofloxacin or meropenem with overall agreement of 91%. The data suggests that XTT is more reliable and accurate than WST-8 for use in a rapid antimicrobial susceptibility test. (c) 2007 Elsevier B.V. All rights reserved.