Classification of functional data with free knots splines
Abstract
Data, obtained through measurements of some process, in many problems, can be treated as functions of a continuous argument. An analysis of such "functional" data is much more complicated than multivariate data analysis. Functional data can be reflected into an appropriate form for traditional statistical algorithms with the help of free -knot splines, which causes almost no loss of information. Finding the free knots of spline is a complex optimization problem, so this paper presents a new heuristic method in order to solve it. An equally important step is to select the parameters of the approximation model. To deal with it, we developed a new approach, which is based on multi-objective optimization of computation time and the accuracy of approximation. The use of splines for classification of functional data was demonstrated on the problem of diagnosis of arthritis based on the bone shapes.
References
Molinari N. Free Knot Splines for Supervised Classifcation // Journal of Classication. — 2007. — № 24. — P. 221–234.
Schwetlick H., and Schutze T. Least squares approximation by splines with free knots // BIT Numerical mathematics. —1995. — № 35. — Р. 361–384.
Yoshimoto F., Harada T., Yoshimoto Y. Data fittingwith a spline using a real-coded algorithm // Journal Computer-Aided Design. — 2003. — № 35. — P. 751–760.
Galvez A., Iglesias A. Efficient particle swarm optimization approach for data fitting with free knot B-splines // Journal Computer-Aided Design. — 2011. — № 43. — P.1683–1692.
Vesterstrom J., Thomsen R. A comparative study of differential evolution, particle swarm optimization, and evolutionary algorithms on numerical benchmark problems // Evolutionary Computation, 2004. CEC2004. Congress on. — 2004. — № 2. — P. 1980–1987.
Ramsay J.O., Silverman B.W. Functional Data Analysis // NY: Springer, 2005.
Ramsay J.O., Silverman B.W. Applied functional data analysis: methods and case studies // New York: Springer-Verlag, 2002. — P. 115–130.
Price K., Storn R.M., Eds. Differential Evolution. A practical approach to global optimization // Germany: Springer-Verlag Berlin, 2005.
Guo Zhenyu, Cheng Bo, Eds. Self-Adaptive Chaos Differential Evolution // Lecture Notes in Computer Science. — 2006. — № 4221. — Р. 972–975.
Pedersen M. Good Parameters for Differential Evolution // Hvass Laboratories, Technical Report no. HL1002. — 2010.
Cramer H. Mathematical Methods of Statistics // Princeton: Princeton University Press, 1946. — P. 282.
