Cluster Analysis of A Tobacco Control Data Set Z. Dzalilov* Graduate School of Information Technology and Mathematical Sciences,University of Ballarat, Victoria, 3353, Australia E-mail Adres: z.dzalilov@ballarat.edu.au A. Bagirov Centre for Informatics and Applied Optimization, University of Ballarat, Victoria, 3353, Australia A B S T R A C T K E Y W O R D S A R T I C L E I N F O Tobacco data set, Global optimization, Cluster analysis, Global k-means algorithm. Received 14 December 2010 Received in revised form 20 December 2010 Accepted 22 December 2010 Available online 23 December 2010 Development of theoretical and methodological frameworks in data analysis is fundamental for modeling complex tobacco control systems. Following this idea, a new optimization based approach was introduced in the paper through two distinct methods: the modified linear least square fit and a heuristic algorithm for feature selection based on optimization techniques. Compared with traditional statistical techniques, optimization- based methods have the potential to detect nonlinearity, and therefore to be more effective analysis tools of complex data set. In this study we evaluate the modified global k-means clustering algorithm by applying it to a massive set of real-time tobacco control survey data. Cluster analysis identified fixed and stable clusters in the studied data. These clusters correspond to groups of smokers with similar behavior and the identification of these clusters may allow us to give recommendations on modification of existing tobacco control systems and on the design of future data acquisition surveys. ________________________________ * Corresponding Author 1. Introduction In this study we evaluate the modified global k-means clustering algorithm by applying it to a tobacco control data set. The International Tobacco Control Policy Evaluation Survey (ITCPES), (ITC survey, 2010) is a recent coordinated international research and evaluation effort. This project provides massive survey data collected from many countries including Australia, for studying and evaluating the psychosocial and behavioral impact of diverse tobacco control policies to smoker behavior across these countries (Figure 1).