Deep Learning Approaches to Improve Effectiveness and Efficiency for Time Series Prediction

The variation and dependency on diverse parameters of time series data make predictions a very complicated process. Also, the effectiveness and efficiency for predicting the values of a time series are important in a variety of areas such as stock exchanges, natural language processing, and sensor networks. Artificial neural networks have been demonstrated to be valuable in such cases to predict the time series data. As recurrent neural network (RNN) is more suitable for predictions of sequential data, we are considering long short-term memory (LSTM) and gated recurrent unit (GRU) in this paper. For comparing with different deep learning frameworks, we use Keras because Keras makes an environment to run new experiments easier and faster. In this paper, we provide a comparative analysis of different deep learning frameworks on the basis of training speed and accuracy of prediction using different metrics, namely mean absolute error (MAE), root mean square error (RMSE), and time taken by different frameworks. The proposed method is expected to be a promising method in the field of time series prediction to choose the suitable deep learning frameworks for a deep learning model.