The packages in this repository can be used to estimate noncrossing conditional quantiles of timeseries using neural networks. We implemented three different models in the Tensorflow framework. This can be found in the tensorflow_NCQRNN package. We implemented one model in the PyTorch framework. This can be found in the pytorch_NCMQRNN package. We also provide a package to test the results. Lastly, we provide a package which can be used to obtain joint prediction region from quantile regression estimates.
See the test file to see how the packages can be used.
Tensorflow version 2.8.0
PyTorch version 1.11.0