Master dissertation, published on SSRN https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4243861
this project code is split into 3 parts: data cleaning+benchmark: the benchmark used for replicate linear hedge fund replication. GAN: all the gan models autoencoders: the actual model is in encapsulate.py, and autoencoder_V4 contain measurement of the result.
directory: data file is the raw data used in data cleaning cleaned data is the output of data cleaning GAN contains all mutations of the model proposed in the paper
I wish i had more time in refining those models, cleaning up and ensemble the prediction/data generation by GAN, but unfortunately it is a master dissertation, i only have 2 months from writing code to writing paper.