Binary Classification System Research Articles

Features Dimensionality Reduction Approaches for Machine Learning Based Network Intrusion Detection

The security of networked systems has become a critical universal issue that influences individuals, enterprises and governments. The rate of attacks against networked systems has increased dramatically, and the tactics used by the attackers are continuing to evolve. Intrusion detection is one of the solutions against these attacks. A common and effective approach for designing Intrusion Detection Systems (IDS) is Machine Learning. The performance of an IDS is significantly improved when the features are more discriminative and representative. This study uses two feature dimensionality reduction approaches: (i) Auto-Encoder (AE): an instance of deep learning, for dimensionality reduction, and (ii) Principle Component Analysis (PCA). The resulting low-dimensional features from both techniques are then used to build various classifiers such as Random Forest (RF), Bayesian Network, Linear Discriminant Analysis (LDA) and Quadratic Discriminant Analysis (QDA) for designing an IDS. The experimental findings with low-dimensional features in binary and multi-class classification show better performance in terms of Detection Rate (DR), F-Measure, False Alarm Rate (FAR), and Accuracy. This research effort is able to reduce the CICIDS2017 dataset’s feature dimensions from 81 to 10, while maintaining a high accuracy of 99.6% in multi-class and binary classification. Furthermore, in this paper, we propose a Multi-Class Combined performance metric C o m b i n e d M c with respect to class distribution to compare various multi-class and binary classification systems through incorporating FAR, DR, Accuracy, and class distribution parameters. In addition, we developed a uniform distribution based balancing approach to handle the imbalanced distribution of the minority class instances in the CICIDS2017 network intrusion dataset.

Docetaxel Treatment in PTEN- and ERG-aberrant Metastatic Prostate Cancers

BackgroundLoss of PTEN is a common genomic aberration in castration-resistant prostate cancer (CRPC) and is frequently concurrent with ERG rearrangements, causing resistance to next-generation hormonal treatment (NGHT) including abiraterone. The relationship between PTEN loss and docetaxel sensitivity remains uncertain. ObjectiveTo study the antitumor activity of docetaxel in metastatic CRPC in relation to PTEN and ERG aberrations. Design, setting, and participantsSingle-centre, retrospective analysis of PTEN loss and ERG expression using a previously described immunohistochemistry (IHC) binary classification system. Patients received docetaxel between January 1, 2006 and July 31, 2016. Outcome measurements and statistical analysisResponse correlations were analyzed using Pearson's χ2 tests and independent-sample t tests. Overall (OS) and progression-free survival (PFS) were analyzed using univariate and multivariate (MVA) Cox regression and Kaplan-Meier methods. Results and limitationsOverall, 215 patients were eligible. Established metastatic CRPC prognostic factors were well balanced between PTEN loss (39%) and normal patients (61%). PTEN loss was associated with shorter median OS (25.4 vs 34.7 mo; hazard ratio [HR] 1.66, 95% confidence interval [CI] 1.18–2.13; p=0.001). There were no differences in median PFS (8.0 vs 9.1 mo; univariate HR 1.20, 95% CI 0.86–1.68; p=0.28) and PSA response (53.4% vs 50.6%; p=0.74). PTEN loss was an independent prognostics factor in MVA. ERG status was available for 100 patients. ERG positivity was not associated with OS or PFS. Limitations include the retrospective nature and the single-centre analysis. ConclusionsOur findings suggest that metastatic CRPC with PTEN loss might benefit more from docetaxel than from NGHT. Patient summaryIn this study we found that metastatic prostate cancer with loss of the PTEN switch may benefit more from docetaxel than from abiraterone.

Classification of medicines using naive bayes classifier

In data mining there is a function called classification that assigns items into groups based on their similarities. These classification techniques are used in many real time applications to classify the data. In pharmaceutical industries all the new medicines need to get approval from Food and Drug administration (FDA) before bringing them into the market. FDA is completely depended on its manual operation and leads to more amount of time for the approval of new medicines. By using classification technique the FDA can automate the procedure with which the time required for approval of new medicine can be reduced. For classification of unknown medicines six chemical properties are used such as relative molecular mass, Number of hydrogen bond donor, Number of hydrogen bond acceptor, Polar surface Area , Hydrophobic constant and number of flexible rotation keys. Drug category and these six properties are collected from U.S National Library of Medicine for creating a new database. By applying Naive Bayes algorithm on the collected data, target class of unknown medicines can be accurately predicted. The fundamental purpose of this paper is to provide a comprehensive analysis of classification methods for predicting unknown medicines. And even discusses the computation procedure of binary classification system for fever and typhoid medicines.

Using generative adversarial networks for improving classification effectiveness in credit card fraud detection

In the last years, the number of frauds in credit card-based online payments has grown dramatically, pushing banks and e-commerce organizations to implement automatic fraud detection systems, performing data mining on huge transaction logs. Machine learning seems to be one of the most promising solutions for spotting illicit transactions, by distinguishing fraudulent and non-fraudulent instances through the use of supervised binary classification systems properly trained from pre-screened sample datasets. However, in such a specific application domain, datasets available for training are strongly imbalanced, with the class of interest considerably less represented than the other. This significantly reduces the effectiveness of binary classifiers, undesirably biasing the results toward the prevailing class, while we are interested in the minority class. Oversampling the minority class has been adopted to alleviate this problem, but this method still has some drawbacks. Generative Adversarial Networks are general, flexible, and powerful generative deep learning models that have achieved success in producing convincingly real-looking images. We trained a GAN to output mimicked minority class examples, which were then merged with training data into an augmented training set so that the effectiveness of a classifier can be improved. Experiments show that a classifier trained on the augmented set outperforms the same classifier trained on the original data, especially as far the sensitivity is concerned, resulting in an effective fraud detection mechanism.

Rotavirus infections in domestic animals

Rotaviruses are major enteric pathogens of humans and a wide variety of animals. Rotavirus infections have a worldwide prevalence. The viral genome is composed of 11 double-stranded RNA segments with six structural and five or six non-structural proteins. Over 35,000 strains have been identified and classified into five main (A, B, C, D, E) and three additional tentative (F, G, H) serotype groups. A binary classification system has been proposed defining ‘G’ or ‘P’ types, with at least 27 G and 35 P genotypes reported thusfar. The virus is transmitted primarily by the faecal-oral and oral routes. After attachment, entry into the host cells occurs through direct entry, fusion or endocytosis. Main mechanisms of Rotavirus-induced diarrhoea involve extensive enterocyte losses and nutrient disdigestion and malabsorption. Clinical features of Rotavirus infections range from asymptomatic infections to fulminant disease leading to rapid death.In calves, lambs and kids, piglets and foals, salient sign of the disease is diarrhoea; diarrhoeic faeces are white, yellow or, in severe cases, blood-tinted or frank haemorrhagic. In dogs and cats, the infection occurs usually as self-limiting diarrhoea. Avian Rotavirus infections are characterized by enteritis, growth depression and/or growth retardation. Definitive diagnosis of the infection can only be achieved by laboratory tests, including electron microscopic examination, immunohistochemical examination, immunofluorescence, ELISA, latex agglutination and molecular techniques. There is no specific treatment against Rotavirus infections. Treatment is based in providing supportive care and managing clinical signs and potential complications. Effective vaccines, containing inactivated, recombinant or attenuated strains of the virus, are available. Challenge studies have shown the ease of the virus in cross-infecting various animal species; animal strains of the virus may cross species and infect humans. Due to the ability of the virus to overcome species barriers, animal strains may act as natural source of viral genomes, promoting mutations and creating new viral genotypes, whose virulence cannot be predictable.

Aggregating validity indicators: The salience of domain specificity and the indeterminate range in multivariate models of performance validity assessment

ABSTRACTThis study was designed to examine the “domain specificity” hypothesis in performance validity tests (PVTs) and the epistemological status of an “indeterminate range” when evaluating the credibility of a neuropsychological profile using a multivariate model of performance validity assessment. While previous research suggests that aggregating PVTs produces superior classification accuracy compared to individual instruments, the effect of the congruence between the criterion and predictor variable on signal detection and the issue of classifying borderline cases remain understudied. Data from a mixed clinical sample of 234 adults referred for cognitive evaluation (MAge = 46.6; MEducation = 13.5) were collected. Two validity composites were created: one based on five verbal PVTs (EI-5VER) and one based on five nonverbal PVTs (EI-5NV) and compared against several other PVTs. Overall, language-based tests of cognitive ability were more sensitive to elevations on the EI-5VER compared to visual-perceptual tests; whereas, the opposite was observed with the EI-5NV. However, the match between predictor and criterion variable had a more complex relationship with classification accuracy, suggesting the confluence of multiple factors (sensory modality, cognitive domain, testing paradigm). An “indeterminate range” of performance validity emerged that was distinctly different from both the Pass and the Fail group. Trichotomized criterion PVTs (Pass-Borderline-Fail) had a negative linear relationship with performance on tests of cognitive ability, providing further support for an “in-between” category separating the unequivocal Pass and unequivocal Fail classification range. The choice of criterion variable can influence classification accuracy in PVT research. Establishing a Borderline range between Pass and Fail more accurately reflected the distribution of scores on multiple PVTs. The traditional binary classification system imposes an artificial dichotomy on PVTs that was not fully supported by the data. Accepting “indeterminate” as a legitimate third outcome of performance validity assessment has the potential to improve the clinical utility of PVTs and defuse debates regarding “near-Passes” and “soft Fails.”

Sentinel node metastasis mitotic rate (SN-MMR) as a prognostic indicator of rapidly progressing disease in patients with sentinel node-positive melanomas.

Risk stratification of sentinel lymph node biopsy (SNB)-positive patients with malignant melanoma differs among current classification systems. To improve classification of patients with rapidly progressive disease who may profit from adjuvant therapy with novel immune or targeted treatment modalities, a single-center retrospective analysis was performed including all melanoma patients diagnosed with a positive SN at a university-based skin cancer center over a 10-year period (2002-2012) (96 of 419 patients). Sentinel node metastasis mitotic rate (SN-MMR) and further histologic parameters were determined by blinded histological re-evaluation and correlated with clinical follow-up (overall [OS], melanoma-specific [MSS], and disease-free survival [DFS]). Median follow-up was 53 months. In univariate analyses, SN tumor penetrative depth (TPD), maximum tumor diameter (MTD), number of positive SN, SN-MMR and the S-, Rotterdam, RDC, Hannover I and II classification systems correlated with OS, MSS and DFS. Multivariate Cox regression analyses showed that a binary classification system based only on the SN-MMR (<1 vs. ≥1 mitoses/mm2 ) was the strongest independent prognostic indicator for all endpoints analyzed. Kaplan-Meier analyses confirmed binary SN-MMR to be superior to stratify patients into high- and low-risk groups (45.45% vs. 87.92% 5-yr MSS). The general prognostic validity of the published SN classification systems was confirmed. The novel SN-MMR classification system may improve discrimination of patients with slowly and rapidly progressive disease. We therefore propose its implementation into clinical practice as the SN-MMR can be easily and reliably determined in routine pathology reports. Its prognostic value for the selection of patients amenable to adjuvant therapies should be studied in clinical trials.

PP - HISTOPATHOLOGICAL AND IMMUNOHISTOCHEMICAL ANALYSIS OF WNT/β-CATENIN SIGNALING PATHWAY PROTEINS IN ACTINIC CHEILITIS

Actinic cheilitis (AC) is a potentially malignant disorder of the lips that is caused by chronic exposure to ultraviolet radiation. The Wnt/β-catenin signaling pathway - also named the canonical signaling pathway - acts in cell proliferation and is involved with tumor development and progression. Some of these molecules have already been identified in AC. Thus, the aim of this study was to comparatively evaluate the histopathological features and the presence of Wnt/β-catenin signaling pathway proteins in AC. For this purpose, immunohistochemical reactions against Wnt1, Wnt5a, β-catenin, axin, APC and cyclin D1 were performed and evaluated in accordance with histopathological epithelial dysplasia grading, using the World Health Organization classification and the binary classification system. The results showed positivity for Wnt 1 (96.7%), cyclin D1 (75.4%), APC (96.7%), axin (95%) and cytoplasmic β-catenin staining (81.9%) and negativity for Wnt 5a (100%). Furthermore, Wnt1, cyclin D1, APC and axin showed an increased labeling index related with the degree of epithelial dysplasia using the binary system. In conclusion, it was possible to state that Wnt 1, cyclin D1 and β-catenin seems to act as tumor promoters by activating the canonical signaling pathway, whereas APC and axin seems to act as tumor promoters by activating the non-canonical signaling pathways.

Evaluation of a Binary Semi-supervised Classification Technique for Probabilistic Record Linkage.

The process of merging data of different data sources is referred to as record linkage. A medical environment with increased preconditions on privacy protection demands the transformation of clear-text attributes like first name or date of birth into one-way encrypted pseudonyms. When performing an automated or privacy preserving record linkage there might be the need of a binary classification deciding whether two records should be classified as the same entity. The classification is the final of the four main phases of the record linkage process: Preprocessing, indexing, matching and classification. The choice of binary classification techniques in dependence of project specifications in particular data quality has not extensively been studied yet. The aim of this work is the introduction and evaluation of an automatable semi-supervised binary classification system applied within the field of record linkage capable of competing or even surpassing advanced automated techniques of the domain of unsupervised classification. This work describes the rationale leading to the model and the final implementation of an automatable semi-supervised binary classification system and the comparison of its classification performance to an advanced active learning approach out of the domain of unsupervised learning. The performance of both systems has been measured on a broad variety of artificial test sets (n = 400), based on real patient data, with distinct and unique characteristics. While the classification performance for both methods measured as F-measure was relatively close on test sets with maximum defined data quality, 0.996 for semi-supervised classification, 0.993 for unsupervised classification, it incrementally diverged for test sets of worse data quality dropping to 0.964 for semi-supervised classification and 0.803 for unsupervised classification. Aside from supplying a viable model for semi-supervised classification for automated probabilistic record linkage, the tests conducted on a large amount of test sets suggest that semi-supervised techniques might generally be capable of outperforming unsupervised techniques especially on data with lower levels of data quality.

Creation of a simple distal femur morphology classification system.

The purpose of this study is to propose a binary classification system based on simple measurements that summarizes individual, race, and sex-specific differences in distal femur shape. Surface models of 165 distal femurs (79 female, 86 male; 85 African-American, 80 Caucasian, 28.8 ± 7.6 years) were created with a 3-dimensional laser scanner. Surface area, width, length, curvature, and angulation were measured. Knees were classified as either type A or B within five distinct categories: (i) aspect ratio, (ii) trochlear-intercondylar width ratio, (iii) trochlear tilt, (iv) medial-lateral trochlear width ratio, and (v) trochlear sulcus shape. Correlations between these measures and surface area were calculated, and receiver-operator curves were used to select cutoff values between type A and B knees to improve differentiation of femur shapes by sex or race. The cutoff values between type A and B knees for the five categories are as follows: Category I: 0.90, Category II: 0.51, Category III: 1.02, Category IV: 0.67, and Category V: 128.7°. Other than category IV (medial-lateral trochlear width ratio) (p = 0.004, R = 0.22), no categories were correlated with surface area (p > 0.25). Category I (aspect ratio, cutoff = 0.90) best differentiated femurs by sex (p < 0.001, AUC = 0.80), and Category V (sulcus shape) best differentiated femurs by race (p < 0.001, AUC = 0.73). This system uses simple measurements to summarize important individual, race, and sex-specific differences in distal femur shape. It can be used in a clinical setting to provide insight into the relationship between sex or race differences in knee shape and mechanically influenced knee disorders. © 2015 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 34:924-931, 2016.

Optimal definition of biological tumor volume using positron emission tomography in an animal model.

BackgroundThe goal of the study is to investigate 18F-fluorodeoxyglucose positron emission tomography (18F-FDG-PET)’s ability to delineate the viable portion of a tumor in an animal model using cross-sectional histology as the validation standard.MethodsSyngeneic mammary tumors were grown in female Lewis rats. Macroscopic histological images of the transverse tumor sections were paired with their corresponding FDG micro-PET slices of the same cranial-caudal location to form 51 pairs of co-registered images. A binary classification system based on four FDG-PET tumor contouring methods was applied to each pair of images: threshold based on (1) percentage of maximum tumor voxel counts (Cmax), (2) percentage of tumor peak voxel counts (Cpeak), (3) multiples of liver mean voxel counts (Cliver) derived from PERCIST, and (4) an edge-detection-based automated contouring system. The sensitivity, which represented the percentage of viable tumor areas correctly delineated by the gross tumor area (GTA) generated from a particular tumor contouring method, and the ratio (expressed in percentage) of the overestimated areas of a gross tumor area (GTAOE)/whole tumor areas on the macroscopic histology (WTAH), which represented how much a particular GTA extended into the normal structures surrounding the primary tumor target, were calculated.ResultsThe receiver operating characteristic curves of all pairs of FDG-PET images have a mean area under the curve value of 0.934 (CI of 0.911–0.954), for representing how well each contouring method accurately delineated the viable tumor area. FDG-PET single value threshold tumor contouring based on 30 and 35 % of tumor Cmax or Cpeak and 6 × Cliver + 2 × SD achieved a sensitivity greater than 90 % with a GTAOE/WTAH ratio less than 10 %. Contouring based on 50 % of Cmax or Cpeak had a much lower sensitivity of 67.2–75.6 % with a GTAOE/WTAH ratio of 1.1–1.7 %. Automated edge detection was not reliable in this system.ConclusionsSingle-value-threshold tumor contouring using 18F-FDG-PET is able to accurately delineate the viable portion of a tumor. 30 and 35 % of Cmax, 30 and 35 % of Cpeak, and 6 × Cliver + 2 × SD are three appropriate threshold values to delineate viable tumor volume in our animal model. The commonly used threshold value of 50 % of Cmax or Cpeak failed to detect one third of the viable tumor volume in our model.

Indicator species for avian biodiversity hotspots: Combination of specialists and generalists is necessary in less natural environments

In this work, I tested the premise that the distribution of a group of few common bird species can be used to predict bird species hotspots in Central Italy. The data on bird observations were collected on 530 sampled sites (150 in cultivated, 150 in forest, 150 in grassland and 80 in urban and peri-urban environments). In each environment, sampled sites with values of bird species richness in the upper than third quartile were classified as high species richness spots (HSRS), while sites with lower bird species richness were classified as non-HSRS (binary classification system).Generalized Linear Models (GLM) were applied using HSRS or non-HSRS as binomial response variable and bird species occurrence was used as the predictor variable. All selected models showed “fair” or “good” capacities to predict the avian hotspots, using only few common birds (4–6) species. However, bird species selected as predictors were different on each environment. In more natural environments (grassland, forest), specialist species were selected, while in most disturbed environments (cultivated and urban) both generalist and specialist species were selected. The results are in agreement with other studies which show how homogenization of bird communities is strongly correlated to landscape disturbance. The findings supports the hypothesis that indicators have to incorporate both specialists and generalist’s species simultaneously. Furthermore, the groups of birds selected as surrogates are easy to detect and this makes it possible to involve citizen-science programmes in obtain data. This approach can be a cheap and efficient and can help to significantly speed up the process of assessing ecosystems that might be under threat.

Clinical implications of genomics for cancer risk genetics

The study of human genetics has provided substantial insight into cancer biology. With an increase in sequencing capacity and a reduction in sequencing costs, genomics will probably transform clinical cancer genetics. A heritable basis for many cancers is accepted, but so far less than half the genetic drivers have been identified. Genomics will increasingly be applied to populations irrespective of family history, which will change the framework of phenotype-directed genetic testing. Panel testing and whole genome sequencing will identify novel, polygenic, and de-novo determinants of cancer risk, often with lower penetrance, which will challenge present binary clinical classification systems and management algorithms. In the future, genotype-stratified public screening and prevention programmes could form part of tailored population risk management. The integration of research with clinical practice will result in so-called discovery cohorts that will help identify clinically significant genetic variation.

Optimizing area under the ROC curve using semi-supervised learning

Receiver operating characteristic (ROC) analysis is a standard methodology to evaluate the performance of a binary classification system. The area under the ROC curve (AUC) is a performance metric that summarizes how well a classifier separates two classes. Traditional AUC optimization techniques are supervised learning methods that utilize only labeled data (i.e., the true class is known for all data) to train the classifiers. In this work, inspired by semi-supervised and transductive learning, we propose two new AUC optimization algorithms hereby referred to as semi-supervised learning receiver operating characteristic (SSLROC) algorithms, which utilize unlabeled test samples in classifier training to maximize AUC. Unlabeled samples are incorporated into the AUC optimization process, and their ranking relationships to labeled positive and negative training samples are considered as optimization constraints. The introduced test samples will cause the learned decision boundary in a multi-dimensional feature space to adapt not only to the distribution of labeled training data, but also to the distribution of unlabeled test data. We formulate the semi-supervised AUC optimization problem as a semi-definite programming problem based on the margin maximization theory. The proposed methods SSLROC1 (1-norm) and SSLROC2 (2-norm) were evaluated using 34 (determined by power analysis) randomly selected datasets from the University of California, Irvine machine learning repository. Wilcoxon signed rank tests showed that the proposed methods achieved significant improvement compared with state-of-the-art methods. The proposed methods were also applied to a CT colonography dataset for colonic polyp classification and showed promising results.11Matlab code of the proposed methods will be released on http://clinicalcenter.nih.gov/drd/summers.html once the paper is published.

An integrated computational approach can classify VHL missense mutations according to risk of clear cell renal carcinoma.

Mutations in the von Hippel–Lindau (VHL) gene are pathogenic in VHL disease, congenital polycythaemia and clear cell renal carcinoma (ccRCC). pVHL forms a ternary complex with elongin C and elongin B, critical for pVHL stability and function, which interacts with Cullin-2 and RING-box protein 1 to target hypoxia-inducible factor for polyubiquitination and proteasomal degradation. We describe a comprehensive database of missense VHL mutations linked to experimental and clinical data. We use predictions from in silico tools to link the functional effects of missense VHL mutations to phenotype. The risk of ccRCC in VHL disease is linked to the degree of destabilization resulting from missense mutations. An optimized binary classification system (symphony), which integrates predictions from five in silico methods, can predict the risk of ccRCC associated with VHL missense mutations with high sensitivity and specificity. We use symphony to generate predictions for risk of ccRCC for all possible VHL missense mutations and present these predictions, in association with clinical and experimental data, in a publically available, searchable web server.

Intraosseous Venous Malformations of the Zygoma

Primary intraosseous venous malformations affecting the zygoma are rare vascular lesions, with only 35 cases reported in the surgical literature. Despite the establishment of the binary classification system, which serves to distinguish vascular tumors from malformations, inappropriate use of the term "hemangioma" to describe a variety of distinct vascular anomalies remains widespread. The authors present 3 cases of zygomatic intraosseous venous malformations and summarize the clinical, radiographic, and immunohistochemical features of these lesions. In each case, an insidious clinical course, combined with the pathognomonic finding of radiating trabeculae on computed tomography, suggests the diagnosis of intraosseous venous malformation. Negative glucose transporter isoform 1 immunoreactivity and histopathological analysis were used to reinforce this diagnosis in 1 patient. Management was individualized in each case, based on symptom complex and aesthetic concern. Given that the therapeutic approach to vascular anomalies is dependent on accurate diagnosis, resolution of semantic matters will optimize the management of these lesions.

Imperviousness Change Analysis Tool (I-CAT) for simulating pixel-level urban growth

Most contemporary urban growth and land cover change models use a binary classification system (e.g. urban/non-urban) to describe urbanized areas. While some models do incorporate continuous urban development indices, they do not specifically provide quantitative information describing the amount of impervious surface within each prediction unit (e.g. pixel or parcel). Such information is essential when characterizing different urban growth patterns, such as high- or low-density development. Here, we present the Imperviousness Change Analysis Tool (I-CAT), which simulates urban growth in terms of pixel-based imperviousness. I-CAT uses common drivers of urban growth (e.g. elevation, slope, distance to water, road networks, and urban centers), along with historic impervious surface area growth trends, to predict the future extent of impervious surface area. A multi-criteria evaluation (MCE) approach is used to determine the respective influence (i.e. weight) of each driver of urban growth. In this study, I-CAT was applied to Jackson County, Missouri. Prediction accuracy, which was assessed by the receiver operating characteristic (ROC) curve method, ranged from 72 to 74 area under curve (AUC). The results demonstrated that I-CAT is an effective tool for providing quantitative estimates of future impervious surface growth.

A novel class dependent feature selection method for cancer biomarker discovery

Identifying key biomarkers for different cancer types can improve diagnosis accuracy and treatment. Gene expression data can help differentiate between cancer subtypes. However the limitation of having a small number of samples versus a larger number of genes represented in a dataset leads to the overfitting of classification models. Feature selection methods can help select the most distinguishing feature sets for classifying different cancers. A new class dependent feature selection approach integrates the F-statistic, Maximum Relevance Binary Particle Swarm Optimization (MRBPSO) and Class Dependent Multi-category Classification (CDMC) system. This feature selection method combines filter and wrapper based methods. A set of highly differentially expressed genes (features) are pre-selected using the F statistic for each dataset as a filter for selecting the most meaningful features. MRBPSO and CDMC function as a wrapper to select desirable feature subsets for each class and classify the samples using those chosen class-dependent feature subsets. The performance of the proposed methods is evaluated on eight real cancer datasets. The results indicate that the class-dependent approaches can effectively identify biomarkers related to each cancer type and improve classification accuracy compared to class independent feature selection methods.

Rediscovering the Contract of Employment for Non-Standard Workers in the UK Common Law

The central concept underlying access to employment rights is based on a binary divide between contract of service (the employee) and contract for services (independent contractor). New classifications of work relationships have however been added over time. These new categories do not sit comfortably with the binary classification system. New ways of working may undermine what we traditionally understand as working relationships. As such the usefulness of this system has been questioned. Yet contract remains central. In view of the lack of any practical alternative it is worth considering whether the interpretation of contract law itself is capable of development. The modern law of contract has the potential capacity to redefine the binary divide so that only those who are truly self-employed are excluded from statutory protection. To do this the meaning of 'contract of service' would need to reflect the reality of labour market contracting and take account of the impact of exploitative bargaining strength. This reinterpretation could potentially expand the classification of contract of service to encompass more non-standard workers within the remit of full employment protection in the UK. Such a possibility depends on how far the employment tribunal and courts have incorporated a contextual framework into their interpretation of contract and how far any context incorporates wider social values consistent with public law standards of fairness and reasonableness. This paper contends that there needs to be explicit acknowledgement by the judiciary that the contextual interpretation of contracts needs to include the economic, political and social imperatives that informed the statutory environment. By seeing the employment relationship in this wider context the pragmatic judge may truly take a purposive stance.

Estimation of a Priori Decision Threshold for Collocations Extraction

Choosing the optimal threshold for the collocations extraction remains a manual task performed by experts. Until today, there is no serious work, based on deep studies, which explores possible solutions to automate the learning of the threshold in the statistical terminology field. In this paper, the authors try to spotlight on this problem by exploring, firstly, the evaluation performance techniques used in several scientific areas (such as biomedical and biometric) and applying them, subsequently, on the statistical terminology field. The experimental study gives promoters results. First, it shows the effectiveness of usual techniques (such as ROC and Precision-Recall curves) used to evaluate the performance of binary classification systems. Second, it provides a practical solution for automatic estimation of optimal thresholds for collocation extraction systems.