Stock Time Series Generation using cGAN
Budget: $80 – $240 HKD
I am following this paper.
https://arxiv.org/pdf/1909.10578v1.pdf
The concept is using 1500 historical trading day data to sample train data which is 400days long, in which 300 days as label and 100 days as prediction. After training is done, I can use the latest 300 days as label and predict 100 future days. then I shift the latest 200 days plus new generated 100 days as label and generate future 200 days, and going on, until I generate future 500 days.
https://arxiv.org/pdf/1909.10578v1.pdf
The concept is using 1500 historical trading day data to sample train data which is 400days long, in which 300 days as label and 100 days as prediction. After training is done, I can use the latest 300 days as label and predict 100 future days. then I shift the latest 200 days plus new generated 100 days as label and generate future 200 days, and going on, until I generate future 500 days.