Tensorflow for Time Series Applications
TimeFlow is a library for building tensorflow models for time series problems. It can be used for time series prediction as well as time series classification. The key features of this library are
- Allows stacking of pre built layers. The layers currently available are
- LSTM(Long Short Term Memory)
- Regression
- Deep Neural networks
- Automated training monitor
- Feature generation for irregular time series
- Other utility functions to facilitate building models
Preinstall Tensorflow before installing TimeFlow. To install the package via github,
git clone https://github.com/abhishekmalali/TimeFlow.git
cd TimeFlow
python setup.py install
###Notes
- Auto correlation reduction network is now included in the package.
- The library works on Tensorflow v1.0