Allan P. White

The Impact of Diversity on Online Ensemble Learning in the Presence of Concept Drift

Online learning algorithms often have to operate in the presence of concept drift (i.e., the concepts to be learned can change with time). This paper presents a new categorization for concept drift, separating drifts according to different criteria into mutually exclusive and nonheterogeneous categories. Moreover, although ensembles of learning machines have been used to learn in the presence of concept drift, there has been no deep study of why they can be helpful for that and which of their features can contribute or not for that. As diversity is one of these features, we present a diversity analysis in the presence of different types of drifts. We show that, before the drift, ensembles with less diversity obtain lower test errors. On the other hand, it is a good strategy to maintain highly diverse ensembles to obtain lower test errors shortly after the drift independent on the type of drift, even though high diversity is more important for more severe drifts. Longer after the drift, high diversity becomes less important. Diversity by itself can help to reduce the initial increase in error caused by a drift, but does not provide the faster recovery from drifts in long-term.

Bias in information-based measures in decision tree induction

A fresh look is taken at the problem of bias in information-based attribute selection measures, used in the induction of decision trees. The approach uses statistical simulation techniques to demonstrate that the usual measures such as information gain, gain ratio, and a new measure recently proposed by Lopez de Mantaras (1991) are all biased in favour of attributes with large numbers of values. It is concluded that approaches which utilise the chi-square distribution are preferable because they compensate automatically for differences between attributes in the number of levels they take.

Technical Note: Bias in Information-Based Measures in Decision Tree Induction

A fresh look is taken at the problem of bias in information-based attribute selection measures, used in the induction of decision trees. The approach uses statistical simulation techniques to demonstrate that the usual measures such as information gain, gain ratio, and a new measure recently proposed by Lopez de Mantaras (1991) are all biased in favour of attributes with large numbers of values. It is concluded that approaches which utilise the chi-square distribution are preferable because they compensate automatically for differences between attributes in the number of levels they take.

Techniques for Dealing with Missing Values in Classification

A brief overview of the history of the development of decision tree induction algorithms is followed by a review of techniques for dealing with missing attribute values in the operation of these methods. The technique of dynamic path generation is described in the context of tree-based classification methods. The waste of data which can result from casewise deletion of missing values in statistical algorithms is discussed and alternatives proposed.

The Importance of Attribute Selection Measures in Decision Tree Induction

Recent work by Mingers and by Buntine and Niblett on the performance of various attribute selection measures has addressed the topic of random selection of attributes in the construction of decision trees. This article is concerned with the mechanisms underlying the relative performance of conventional and random attribute selection measures. The three experiments reported here employed synthetic data sets, constructed so as to have the precise properties required to test specific hypotheses. The principal underlying idea was that the performance decrement typical of random attribute selection is due to two factors. First, there is a greater chance that informative attributes will be omitted from the subset selected for the final tree. Second, there is a greater risk of overfitting, which is caused by attributes of little or no value in discriminating between classes being “locked in” to the tree structure, near the root. The first experiment showed that the performance decrement increased with the number of available pure-noise attributes. The second experiment indicated that there was little decrement when all the attributes were of equal importance in discriminating between classes. The third experiment showed that a rather greater performance decrement (than in the second experiment) could be expected if the attributes were all informative, but to different degrees.

Website design attributes for retrieving health information by older adults: an application of architectural criteria

The objectives of this research were to identify design attributes to develop easy-to-use websites for older adults. Forty-one males and 58 females (age range 58–90) were asked to retrieve information on a health-related topic from the NHS Direct and Medicdirect websites, and were asked to fill in a website evaluation questionnaire. An exploratory factor analysis of data identified navigation/search usability, link usability, usefulness and colour as important dimensions of a senior-friendly website. A two-stage, three-component regression model with these dimensions as predictor variables and the satisfaction level in using a website as the dependent variable has been proposed.

An adoption study comparing the prevalence of psychiatric illness in women who have adoptive and natural children compared with women who have adoptive children only

The current study compares the current and lifetime prevalence of affective disorder in women who have adopted and have natural children (n = 110) with women who only have adopted children (n = 176). There was no difference in lifetime prevalence of psychiatric disorder between the two groups and a nonsignificant trend for women who had born children to have had a major depressive episode during their lifetime 48 (44%) cf 62 (35%). The increased prevalence of psychiatric illness in married women with children cannot be explained by the biological fact of bearing children. None of the social variables related to child-rearing which were examined influenced the lifetime prevalence of psychiatric disorder.

Machine learning techniques in early screening for gastric and oesophageal cancer

A database on 2692 dyspeptic patients over the age of 40 was established, consisting of 73 epidemiological and clinical variables. A tree-based machine learning algorithm (PREDICTOR) was applied to this database, in order to attempt to find rules which would classify patients into 2 groups, i.e., those suffering from gastric or oesophageal cancer, and the remainder. The results were encouraging. The cross-validated classification performance figure showed that by classifying 61.3% of the patients as high risk, a sensitivity of 94.9% and a specificity of 39.8% could be achieved. It is planned to construct an expert system based on the rules produced by the machine learning algorithm, in order to provide preliminary screening for cancer in dyspeptic patients.

Effects of playing a wind instrument on the occlusion.

INTRODUCTION There is a popular belief among some musicians that playing a wind instrument regularly can affect the position of the teeth. The aim of this study was to investigate this hypothesis. METHODS A cross-sectional observational study was carried out, comparing the occlusions of 170 professional musicians selected from 21 orchestras and organizations in the United Kingdom. The subjects were subdivided according to type of instrument mouthpiece and included 32 brass players with large cup-shaped mouthpieces, 42 brass players with small cup-shaped mouthpieces, and 37 woodwind players with single-reed mouthpieces. Fifty-nine string and percussion players formed the control group. Impressions were taken of the teeth of each subject, and occlusal parameters were assessed from the study casts. The results were analyzed by using analysis of variance (ANOVA) and chi-square tests. RESULTS No statistically significant differences were found in overjet (P = 0.75), overbite (P = 0.55), crowding (maxillary arch, P = 0.31; mandibular arch, P = 0.10), irregularity index (maxillary arch, P = 0.99; mandibular arch, P = 0.16), and the prevalence of incisor classification (P = 0.15) between the wind instrument players and the control group. However, the large-mouthpiece brass group had a significantly higher prevalence of lingual crossbites in comparison with all other groups. CONCLUSIONS Playing a wind instrument does not significantly influence the position of the anterior teeth and is not a major etiologic factor in the development of a malocclusion. However, playing a brass instrument with a large cup-shaped mouthpiece might predispose a musician to develop lingual crossbites or lingual crossbite tendencies.

A twin study examining the effect of parity on the prevalence of psychiatric disorder

The study compares the prevalence of current and lifetime psychiatric illness in twin pairs where one twin is more parous than the other. The main finding was that parity did not affect the prevalence of psychiatric disorder. This adds confirmation to the study of Bebbington et al. (1991) that high prevalence rates in women with children is due to an effect of marriage rather than an effect of parity. There was a diagnostic difference between the twins with the less parous twin having significantly more cases of Research Diagnostic Criteria (RDC; Spitzer et al., 1978) major and minor depression and the more parous twin more cases of generalised anxiety disorder. The life events and difficulties in the previous 12 months were also established in the whole group and events and difficulties over the whole of the subjects life for the first 20 pairs. The total number of events in the previous 12 months was significantly higher in the more parous twin of the pair, these were child-related events and not associated with an increased likelihood of current psychiatric illness. There were no differences in the number of events and difficulties during the lifetime between the first 20 twin pairs.