Link to Codalab competition.
The aim of this project to predict future samples of a multivariate time series. The goal is to design and implement forecasting models to learn how to exploit past observations in the input sequence to correctly predict the future.
The data can be found under the link. The provided time series have a uniform sampling rate.
- Length of the time series (number of samples in the training set): 68528.
- Number of features: 7.
- Name of the features: 'Sponginess', 'Wonder level', 'Crunchiness', 'Loudness on impact', 'Meme creativity', 'Soap slipperiness', 'Hype root'.
- Convolutional Bidirectional LSTM
- Convolutional Bidirectional LSTM using Bahdanau's Attention
- Convolutional Bidirectional LSTM using Luong's Attention
- Seq2Seq LSTM using Luong's Attention
- Stacked LSTM
- Stacked SCINet