Idempotent linearly compensated hybrid connectives for customers patterns segmentation

Abstract. When employing unsupervised learning methods to auto- matically generate segmentations from the available data, often a lot of alternative segmentations are produced. In this paper the unsuper- vised learning capability of LAMDA (Learning Algorithm for Multivari- ate Data Analysis) algorithm is applied and the results are intended to help marketing experts to ¯nd out hidden patterns among their cus- tomers pro¯les. LAMDA algorithm, uses hybrid linearly compensated connectives in the learning process. The condition of idempotence imposed to the hybrid connectives reduces the space of search and guaranties coherence in the obtained segmentation. The obtained results are applied to identify the behaviour and needs of the customers of a Spanish retailer.