Enhancing assessment of Personalized Multi-Agent System through ConvLSTM

Abstract

Personalized system represents nowadays a key factor to attract user and to guide special end users concerns. All systems that include interaction with end users are seeking to improve this factor, particularly, Personalized systems that use Multi-Agent System (PMAS), characterized by agents collaboration and parallel execution of tasks, as a tool to implement the personalization. However, assessing this kind of personalized system is complicated and generally restrained to empirical assessment. In this context, our purpose is to process the assessment of PMAS through the confrontation between the system to be evaluated and a referential one. In this paper, we deal with the web environment. We describe first our referential system, then, we focus on the prediction layer of this system. We demonstrate the use of Convolutional Long Short-Term Memory (ConvLSTM), a machine learning approach, to treat the semantic reasoning of the end users.