Open Access Journal

ISSN : 2394-2320 (Online)

International Journal of Engineering Research in Computer Science and Engineering (IJERCSE)

Monthly Journal for Computer Science and Engineering

Open Access Journal

International Journal of Engineering Research in Computer Science and Engineering (IJERCSE)

Monthly Journal for Computer Science and Engineering

ISSN : 2394-2320 (Online)

Classification of Time Series Trajectories Based on Shape Features using SVM

Author : Gajanan S. Gawde 1 Jyoti D. Pawar 2

Date of Publication :7th March 2017

Abstract: Classification of time series trajectories is a non trivial problem and few researchers have contributed to this field. Time series trajectories such as Character trajectories, ECG trajectories, Stock Exchange Trajectories need to be classified accurately in order to improve the performance of system. The classification of time\ series trajectories is carried out using different techniques such as Neural networks, Support vector machines, Fourier Transforms and Bayesian Network. Most of the existing classification techniques work on simple features such as co-ordinates, distance, and velocity. These techniques are not able to consider shape features of trajectories and classify the test trajectories. Shape of the time series trajectory is an important feature and can be used in the classification method. A support vector machine is better classification technique compared to the existing techniques. In this paper, we have proposed our novel classification method to classify time series trajectories using SVM with shape as the feature vector for classification purpose. Thorough experimental study was carried out on proposed technique with different datasets. Experimental results shows that SVM technique is better compared to Neural Network, Fourier Transform, Bayesian Network. SVM with shape features is efficient compared to SVM without shape features. We have tested SVM with different kernel functions. Our experimental results show that RBF SVM with shape features method is efficient compared with Linear and Non Linear kernels.

Reference :

Will Updated soon

Recent Article