GDTW-P-SVMs: Variable-length time series analysis using support vector machines

摘要: We describe a new technique for sequential data analysis, called GDTW-P-SVMs. It is maximum margin method the construction of classifiers with variable-length input series. employs potential support vector machines (P-SVMs) and Gaussian Dynamic Time Warping (GDTW) to waive fixed-length restriction feature vectors in training test data. As result, GDTW-P-SVMs enjoy P-SVM method's properties such as ability to: (i) handle kernel matrices that are neither positive definite nor square (ii) minimise scale-invariant capacity measure. The elaborates on functions, by utilising well-known dynamic time warping algorithm provide an elastic distance measure functions. Benchmarks classification performed several real-world sets from UCR series classification/clustering page, GeoLife trajectory set, UCI Machine Learning Repository. include both variable results show performs significantly better than benchmarked standard methods.