The Combination of Multiple Classifiers Using an Evidential Reasoning Approach

In many domains when we have several competing classifiers available we want to synthesize them or some of them to get a more accurate classifier by a combination function. In this paper we propose a ‘class-indifferent’ method for combining classifier decisions represented by evidential structures called triplet and quartet, using Dempster's rule of combination. This method is unique in that it distinguishes important elements from the trivial ones in representing classifier decisions, makes use of more information than others in calculating the support for class labels and provides a practical way to apply the theoretically appealing Dempster–Shafer theory of evidence to the problem of ensemble learning. We present a formalism for modelling classifier decisions as triplet mass functions and we establish a range of formulae for combining these mass functions in order to arrive at a consensus decision. In addition we carry out a comparative study with the alternatives of simplet and dichotomous structure and also compare two combination methods, Dempster's rule and majority voting, over the UCI benchmark data, to demonstrate the advantage our approach offers. (A continuation of the work in this area that was published in IEEE Trans on KDE, and conferences)

Towards optimal clinical and epidemiological registration of haematological malignancies: Guidelines for recording progressions, transformations and multiple diagnoses

Haematological malignancies (HM) represent over 6% of the total cancer incidence in Europe and affect all ages, ranging between 45% of all cancers in children and 7% in the elderly. Thirty per cent of childhood cancer deaths are due to HM, 8% in the elderly. Their registration presents specific challenges, mainly because HM may transform or progress in the course of the disease into other types of HM. In the context of cancer registration decisions have to be made about classifying subsequent notifications on the same patient as the same tumour (progression), a transformation or a new tumour registration. Allocation of incidence date and method of diagnosis must also be standardised. We developed European Network of Cancer Registries (ENCR) recommendations providing specific advice for cancer registries to use haematology and molecular laboratories as data sources, conserve the original date of incidence in case of change of diagnosis, make provision for recording both the original as well as transformed tumour and to apply precise rules for recording and counting multiple diagnoses. A reference table advising on codes which reflect a potential transformation or a new tumour is included. This work will help to improve comparability of data produced by population-based cancer registries, which are indispensable for aetiological research, health care planning and clinical research, an increasing important area with the application of targeted therapies.

Are food-related perceptions associated with meal portion size decisions? A cross-sectional study

The purpose of this study was to test a comprehensive model of meal portion size determinants consisting of sociodemographic, psychological and food-related variables, whilst controlling for hunger and thirst.
Using cross-sectional nationally representative data collected in 2075 participants from the Island of Ireland (IoI) and Denmark (DK), eight separate hierarchical multiple regression analyses were conducted to examine the association between food-related variables and meal portion size (i.e. pizza, vegetable soup, chicken salad and a pork meal) within each country. Stepwise regressions were run with physiological control measures (hunger and thirst) entered in the first step, sociodemographic variables (sex, age, body mass index (BMI)) in the second step; psychological variables (cognitive restraint, uncontrolled eating, emotional eating, general health interest (GHI)) in the third step and food-related variables (expected fillingness, liking, expected healthfulness, food familiarity) in the fourth step.
Sociodemographic variables accounted for 2-19% of the variance in meal portion sizes; psychological variables explained an additional 3-8%; and food-related variables explained an additional 2-12%. When all four variable groups were included in the regression models, liking and sometimes expected healthfulness was positively associated with meal portion size. The strongest association was for liking, which was statistically significant in both countries for all meal types. Whilst expected healthfulness was not associated with pizza portion size in either country, it was positively associated with meals that have a healthier image (vegetable soup; chicken salad and in IoI, the pork meal).
In conclusion, after considering sociodemographic and psychological variables, and the food-related variables of liking and expected healthfulness, there may be little merit in manipulating the satiating power, at least of these type of meals, to maintain or promote weight loss.
Keywords: Meal portion size; psychological variables; expected fillingness; expected healthfulness; food liking; food familiarity.