Predictive data mining driven architecture to guide car seat model parameter initialization

Abstract

Researchers in both government and non government organizations are constantly looking for patterns among drivers that may influence proper use of car seats. Such patterns will help them predict behaviours of drivers that shape their decision in placing a child in the proper car constraint when traveling in an automobile. Previous work on a multi-agent based prototype, with the goal to simulate car seat usage patterns among drivers, has shown good prospects as a tool for researchers. In this work we aim at exploring the parameters that initialize the model. The complexity of the model is driven by a large number of parameters and a wide array of values. Existing data from road surveys are examined using existing data mining tools in order to explore beyond basic statistics what parameters and values can be most relevant for a more realistic model run. The intent is to make the model replicate real world conditions as closely mimicking the survey data as possible. Data mining driven architecture which can dynamically use data collected from various surveys and agencies in real time can significantly improve the quality and accuracy of the agent-model. © 2011 Springer-Verlag Berlin Heidelberg.