TensorFlow-Keras implementation for ForGAN
This is a tensorflow-keras implementation for the Forecasting GAN (ForGAN) which was originally implemented by Koochali et. al (https://arxiv.org/pdf/1903.12549v1.pdf).
GAN was originally implemented to handle the univariate analysis and prediction, with very few modifications I have tried to do multi-variate analysis as well.
The Data used is the Stock data of GE company available in Kaggle. I have uploaded the the data sample as well.
components.py makes compoments of the GANs such as discriminator, generator and also a combinational network which is required to train the generator.
utils,py contains crutial functions required for KLD calculation and data input.
mutlistep.py has some adaptation from original code and it can be used to do multi-varaiate / uni-variate analysis and prediction as well.
I have also included few of the graph showing the prediction of the network.