A Multi-agent Architecture for Labeling Data and Generating Prediction Models in the Field of Social Services

Abstract

Prediction models are widely used in insurance companies and health services. Even when 120 million people are at risk of suffering poverty or social exclusion in the EU, this kind of models are surprisingly unusual in the field of social services. A fundamental reason for this gap is the difficulty in labeling and annotating social services data. Conditions such as social exclusion require a case-by-case debate. This paper presents a multi-agent architecture that combines semantic web technologies, exploratory data analysis techniques, and supervised machine learning methods. The architecture offers a holistic view of the main challenges involved in labeling data and generating prediction models for social services. Moreover, the proposal discusses to what extent these tasks may be automated by intelligent agents.