Abstract—Time series classification (TSC) has become a
popular research topic in recent years. The TSC works with any
real value sequences such as regular discrete time signals as well
with time series obtained by conversion from shapes in an
image. Recently, time series shapelets classification methods
have been used for data mining and time series classification.
These methods are proven to have high accuracy and
outperform state-of-the-art methods such as the nearest
neighbor in cases where the signals are noisy. Furthermore
these methods work well with local regions of the signal instead
with the whole signal, a technique that in some cases gives better
results as the global features are more susceptible to distortions.
However, one of the main drawbacks of the time series shapelets
classification methods is its slow training time. In this paper, we
propose a method that is based on particle swarm optimization
(PSO) to improve not only the training time but also
performance accuracy when applied to benchmark datasets.
Index Terms—Time series shapelets, particle swarm
optimization (PSO), classification.
The authors are with the Mississippi State University, Mississippi State,
MS 39762, USA (e-mail: ism6@msstate.edu, younan@ece.msstate.edu).
Cite: Ivan S. Mitzev and Nickolas H. Younan, "Time Series Shapelets: Training Time Improvement Based on Particle Swarm Optimization," International Journal of Machine Learning and Computing vol. 5, no. 4, pp. 283-287, 2015.