B0692
Title: Projection pursuit: An empirical application to Italian primary school children
Authors: Cinzia Franceschini - Bologna University (Italy) [presenting]
Nicola Loperfido - University of Urbino (Italy)
Abstract: The University of Gastronomic Sciences (Pollenzo, Italy) investigated the attitude of Italian children towards food and its consumption in school canteens. Data were collected from questionnaires administered to 1108 children in 9 primary Italian schools. Original data is first clustered by means of model-based clustering and k-means clustering. Then, principal component analysis is used to reduce the number of variables before clustering. The best clustering is obtained using k-means on the data projected onto the directions found using projection pursuit, a multivariate statistical technique aimed at finding interesting low-dimensional data projections. Projection pursuit addresses three major challenges of multivariate analysis: the curse of dimensionality, irrelevant features and the limitations of visual perception. The data at hand makes a case for projection pursuit for variable reduction within a classification framework, even when the number of variables is much smaller than the number of units.
