Analysis of various deep learning based models for financial time series data using convolutions, recurrent neural networks (lstm), dilated convolutions and residual learning
Dow Jones Industrial Average stock data downloaded from Yahoo finance from the year 2000- 2017
Compared the various existing approaches to time series data analysis by using proposed architectures based on convolutional neural networks, recurrent neural networks or LSTM's, dilated convolutions and residual neural networks based on the current Wavenet architecture. The current state of the art approach is combines with an LSTM and basic convolutions to obtain results that surpass the current state of the art approaches.
execute the files stock_conv1d.ipynb, stock_dilatedconv+lstm.ipynb, stock_lstm+conv1d.ipynb, stock_nn.ipynb, stock_rnn_lstm.ipynb and compare the results