Ensemble Deep TimeNet: An Ensemble Learning Approach with Deep Neural Networks for Time Series

Time series classification (TSC) is the problem of predicting class labels of time series generated by different signal sources. TSC has been a challenging problem in machine learning and statistics for many decades. TSC has many important applications in bioinformatics, biomedical engineering, and clinical predictions. A large number of classification algorithms have been developed to address TSC problem. However, there is still a lot of room for improving the accuracy of classification. Traditional approaches extract discriminative features from the time series data by applying different types of transformations. These features are then fed into standard classifiers for classification. After a tremendous success of deep neural networks in certain areas such as NLP, image processing, and speech recognition, some researchers have applied deep convolutional neural networks and recurrent neural network based approaches for TSC. Deep neural network based algorithms have established a new baseline for TSC. In this paper, we propose Ensemble Deep TimeNet (EDTNet), an ensemble of multiple deep neural networks for TSC. We have compared the accuracy of EDTNet with those of the state-of-the-art algorithms on 44 different datasets from UCR time series database. Through extensive experiments we show that EDTNet outperforms all the competing algorithms in most of the UCR datasets in terms of classification accuracy. Specifically, EDTNet outperforms the other algorithms on 26 out of the 44 datasets, whereas, the next best algorithm (FCN) has a better accuracy than the others in only 10 out of the 44. A GPU based python implementation of our algorithm will be posted at https://github.com/sudiptap/Timeseries-Classification when the paper is accepted.