Hathaway and Bezdek's Fuzzy c-Regression Models (FCRM) is regarded as a simultaneous analysis of clustering and multiple regression. In high dimensional setting, the regression techniques do not perform well for reasonable sample sizes because of the inherent sparsity of samples in the cluster. This paper proposes a simultaneous approach to the clustering, principal component analysis and multiple regression analysis. Aiming at knowledge discovery from POS data that are recorded automatically into the cash register transactions, we analyze the relationship between meteorological information and sales of perishables.
