Hybrid Neural Networks for Learning the Trend in Time Series

Trend of time series characterizes the intermediate upward and downward behaviour of time series. Learning and forecasting the trend in time series data play an important role in many real applica- tions, ranging from resource allocation in data cen- ters, load schedule in smart grid, and so on. In- spired by the recent successes of neural networks, in this paper we propose TreNet, a novel end-to- end hybrid neural network to learn local and global contextual features for predicting the trend of time series. TreNet leverages convolutional neural net- works (CNNs) to extract salient features from local raw data of time series. Meanwhile, considering the long-range dependency existing in the sequence of historical trends, TreNet uses a long-short term memory recurrent neural network (LSTM) to cap- ture such dependency. Then, a feature fusion layer is to learn joint representation for predicting the trend. TreNet demonstrates its effectiveness by out- performing CNN, LSTM, the cascade of CNN and LSTM, Hidden Markov Model based method and various kernel based baselines on real datasets.