Evaluation of the effectiveness of features selected by discriminant analysis methods

New theoretical and practical results concerning the use of discriminant analysis for feature selection are presented in the paper. Numerical values for the eigenvalues of the matrix SW_W⁻¹ S_B (within-class and between-class scatter matrices) are investigated. An analytic expression for their minimum value representing the minimum effectiveness is derived. Differences between real values and these minimum values are important for the evaluation of the effectiveness of features and thus for feature selection.     

Fast branch & bound algorithms for optimal feature selection

A novel search principle for optimal feature subset selection using the Branch & Bound method is introduced. Thanks to a simple mechanism for predicting criterion values, a considerable amount of time can be saved by avoiding many slow criterion evaluations. We propose two implementations of the proposed prediction mechanism that are suitable for use with nonrecursive and recursive criterion forms, respectively. Both algorithms find the optimum usually several times faster than any other known Branch & Bound algorithm. As the algorithm computational efficiency is crucial, due to the exponential nature of the search problem, we also investigate other factors that affect the search performance of all Branch & Bound algorithms. Using a set of synthetic criteria, we show that the speed of the Branch & Bound algorithms strongly depends on the diversity among features, feature stability with respect to different subsets, and criterion function dependence on feature set size. We identify the scenarios where the search is accelerated the most dramatically (finish in linear time), as well as the worst conditions. We verify our conclusions experimentally on three real data sets using traditional probabilistic distance criteria.     

Filter‐ versus wrapper‐based feature selection for credit scoring

We address the problem of credit scoring as a classification and feature subset selection problem. Based on the current framework of sophisticated feature selection methods, we identify features that contain the most relevant information to distinguish good loan payers from bad loan payers. The feature selection methods are validated on several real‐world datasets with different types of classifiers. We show the advantages following from using the subspace approach to classification. We discuss many practical issues related to the applicability of feature selection methods. We show and discuss some difficulties that used to be insufficiently emphasized in standard feature selection literature. © 2005 Wiley Periodicals, Inc. Int J Int Syst 20: 985–999, 2005.     

Determination of aryloxyphenoxypropanoic acid derivatives in crops treated with herbicidal sprays

Summary A reproducible and sensitive gas Chromatographic method has been developed for the trace analysis of fluazifop-butyl, haloxyfop-ethoxyethyl and quizalofop-ethyl as well as their main metabolites (corresponding acids analysed after conversion to methyl esters) in some fruits and vegetables treated with herbicidal sprays. Comparable detection limits (0.01 mg/kg) were achieved when mass fragmentography and/or the electron capture detector were employed. However, in the latter case, bromination of fluazifop esters had to be carried out prior to GC analysis. By means of the nitrogen-phosphorus detector, determination of residues at a level of 0.05 mg/kg was possible. The residue values found in strawberries, cabbages, cauliflowers and carrots were used to discuss the degree of metabolic hydrolysis of parent esters, i.e. active ingredients of herbicides.

---

Bestimmung von Aryloxyphenoxypropionsäure-Derivaten in mit Herbicidsprays behandelten Produkten

Zusammenfassung Es wurde eine gaschromatographische Methode zur quantitativen und selektiven Bestimmung von Fluazifop-butyl, Haloxyfop-ethoxyethyl und Quizalofop-ethyl sowie von ihren Hauptmetaboliten (den entsprechenden Säuren, die zu Methylestern umgesetzt werden müssen) in mit Herbiciden behandelten Obst- und Gemüseproben entwickelt. Eine Nachweisgrenze von 0,01 mg/kg wurde durch Elektroneneinfanggaschromatographie oder durch Massenfragmentographie erzielt. Um die gleiche Empfindlichkeit des EC-Detektors bei Fluazifop zu erreichen, wurden die beiden Ester mit Brom umgesetzt. Mit Hilfe des NP-Detektors konnte eine Nachweisgrenze von 0,05 mg/kg erreicht werden. Die festgestellten Rückstandsgehalte in Erdbeeren, Kohl, Blumenkohl und Möhren wurden gleichzeitig zur Beurteilung der metabolischen Hydrolyse der diskutierten Ester (der Wirkstoffe der Herbicide) herangezogen.

     

Some substrates and inhibitors of cytosolic epoxide hydrolase induce sister-chromatid exchanges in mammalian cells, but do not induce gene mutations in Salmonella typhimurium and V79 cells

We describe a 66-year-old man who presented initially with acute cholecystitis. He was treated by cholecystostomy and biopsy of the gallbladder mucosa which revealed carcinoma of the gallbladder. Four weeks later a cholecystectomy was performed followed by resection of the common bile duct, common hepatic duct and segments IV and V of the liver and a hepaticojejunostomy. Sixteen months later an abdomino-perineal resection was performed for a moderately differentiated Dukes' stage C carcinoma of the rectum. He is alive and without evidence of recurrence seven years later. Few patients survive for this length of time following resection of either carcinoma of the gallbladder or rectum. This case report demonstrates the value of aggressive surgical treatment in patients with early carcinoma of the gallbladder.     

Traceability of olive oil based on volatiles pattern and multivariate analysis

An automated head-space solid-phase microextraction (HS-SPME)-based sampling procedure, coupled to gas chromatography–ion trap mass spectrometry (GC–ITMS), was developed and employed for fast characterisation of olive oil volatiles. In total, 914 samples were collected, over three production seasons, in north-western Italy—Liguria (n = 210) and other regions—in addition to the rest of Italy, Spain, France, Greece, Cyprus, and Turkey (n = 704) with the aim to distinguish, based on analytical (profiling) data, the olive oils labelled as “Ligurian” (protected denomination of origin region, PDO) from all the others (“non-Ligurian”). For the chemometric analysis, linear discriminant analysis (LDA) and artificial neural networks with multilayer perceptrons (ANN-MLP) were tested. Employing LDA, somewhat lower recognition (81.4%) and prediction (61.7%) abilities were obtained. The classification model was significantly improved using ANN-MLP. Under these conditions, the recognition (90.1%) and prediction (81.1%) abilities were achieved. The diagnostic value of the data obtained by one-dimensional GC–ITMS were compared with those generated by two-dimensional gas chromatography–time-of-flight mass spectrometry (GC × GC–TOFMS), allowing a comprehensive analysis of olive oil volatiles.     

The thermostability of paperboard food packing

Three different methods of isolation of volatiles released by heating of the paperboard were compared. The volatiles produced by 30–60 min of heating at 150 °C were isolated by: 1) extraction with diethyl ether 2) distillation with trapping in cooled diethyl ether and 3) adsorption on TENAX GC. The isolated volatiles were analyzed by gas chromatography coupled with the mass spectroscopy detection. The TENAX adsorption seems to give the best results, better than the other methods, by use of which not only volatiles were isolated and some part of volatiles was lost also by subsequent reactions of those.     

Residues of clopyralide (3,6-dichloropicolinic acid) in strawberries and their changes

Residues of clopyralide in strawberries treated with herbicide preparation Lontrel 300 were determined using gas chromatographic method. The decrease of clopyralide content during the processing of contaminated fruits was observed. The rate of decomposition of this compound in model solutions (held at 75 degrees C, pH 3.6) followed first-order kinetics. 2,5-dicloropyridine was proved to originate under these conditions from parent compound.     

Prediction of odour and flavour acceptances of white wines on the basis of their colour

The colour of Moravian white wines was evaluated instrumentally, using the trichromatic method, and by sensory analysis. The odour and flavour acceptances could be predicted very well on the basis of sensory colour evaluation, and less efficiently by instrumental method. Consumers preferred wines with prevailing yellow hue, which was associated with sweet, fruity and floral flavour notes. Green hue was considered a negative factor, without any rational relation to negatively perceived flavour notes. Logarithmic relations fit moderately better experimental results than linear relations.     

Association of class I HLA antigens with invasive meningococcal disease

BACKGROUND: The majority of meningococcal infections are characterized by nasopharyngeal carriership. In some patients invasive disease with a mild course develops, while some cases have a lethal outcome. The reasons of this wide variation range are not clear. The objective of the present work was to assess whether the development of invasive meningococcal disease or its prognosis are associated with HLA class I. METHODS AND RESULTS: The group of patients was formed by 40 patients (29 females, 11 males, mean age 16 years, range 8 months to 52 years). In 28 patients the disease was caused by N. meningitidis group C, in 9 cases group B, in three cases the serotype was not assessed. The etiology was confirmed by cultivation or latex agglutination. Twenty-three patients had a mild course of the disease, 8 a medium severe one, 9 patients a severe clinical course (score according to Stiehme, Damrosch and Rosenblat). The patients were compared with 227 non-related blood donors (114 women, 113 men, 18 to 50 years old). In patients and controls 24 lymphocytic HLA antigens class I were identified as to type. Typing was done using the standard microlymphocytotoxic test in the NIH modification. The results were processed by statistical methods using Fisher's exact test and the 2 x 2 test with Yates correction. In patients with a mild course HLA antigens B7 and B12 predominate (p = 0.03; p = 0.02), in medium severe cases antigen A11 (p = 0.03), in patients with the most severe course antigen A9 (p = 0.04). In invasive infections caused by N. meningitidis serotype B antigen B17 predominates (p = 0.05). CONCLUSIONS: The severity of meningococcal invasive infections is associated with HLA class I. Invasive disease caused by N. meningitidis serotype B are more likely to occur in carriers of HLA B17. No relationship was found between HLA class I and invasive disease caused by N. meningitidis regardless of serotype and with serotype C.