Perfiles de usuario para agentes de interfaz: un análisis de técnicas de aprendizaje

Abstract

Agents that help users when they buy products on the Web promise to be a suitable way for transforming the overloaded world of web offers into a nice shop with personalized assistance. This promise depends closely on the efficiency of these agents to determine the preferences and habits of their customers. In this context, we have built an agent named Vendor that interacts with web customers. Through this agent we analyze three of the machine learning techniques that present advantageous characteristics for building user profiles. Thus, Bayesian networks, genetic algorithms, and neural networks have been compared for customer profiling.