Skip to content

Applications of optimization based data mining technique to a tobacco control data set.

Zari Dzalilov - 06/10/2010

Speaker:

Zari Dzalilov
Abstract:

Control of tobacco smoking is a global health issue as the use of cigarettes is recognized to be the second leading cause of death around the world.

Tobacco control may depend on the development of a range of new theoretical and methodological frameworks for describing and understanding complex tobacco control systems. Because the relationships between outcomes of tobacco control and many key policies are often nonlinear, optimization-based methods have the potential to be more effective tools of complex tobacco control systems, when compared to traditional statistical techniques. In this study, we evaluate two global optimization-based algorithms that determine the contribution of multiple features to the effectiveness of outcome (such as successful quitting). These algorithms are the modified linear least square fit and a heuristic algorithm for feature selection based on optimization techniques. These methods explore the relationship between features and classes and allow the use of data sets with an arbitrary number of classes. Preliminary results show a number of significant features associated with quitting attempts. We are also considering implementations of different scenarios, based on combinations of different optimization algorithms. 
 
 

Related links