Abstract:
Moving objects generate huge number of time series trajectories and length of such time series trajectories are large. To process such lengthy time series trajectories is very time consuming. There is need to reduce the dimension of the time series trajectory so that overall execution time would get reduced. There are few techniques proposed to tackle dimension reduction issue and none of them are working on a shape feature of the time series trajectory. PCA and SVM techniques are used to reduce the time series trajectories. This methods are not considering shape of the trajectories while reduction. In this paper, we have proposed Shape_PCA dimension reduction technique using Principle Component Analysis with shape as a feature vector. Our proposed technique identifies shapes which are repeated more number of times in the trajectory and accordingly reduce the time series trajectory. Thorough experimental study was carried out with different datasets. Shape_PCA method was compared with PCA and SVM without shape feature. The experimental results show that, Shape_PCA technique reduction ratio is higher side compared to PCA and SVM method. Thus, Shape_PCA is efficient method compared to PCA and SVM technique.
