Moving window size for time series analysis : Minimum description length and Fractal data dimension

I'm doing the time-series analysis as a hobby and am attempting to discover the optimal window size. After some study, I found 2 techniques that had been mentioned on the internet:

1. MDL (minimum descriptive length)

2. Fractal data dimension : http://www.mental.sk/temp/ai/tsp.html

I am starting to develop a python program to accomplish these 2 approaches. I do know about MDL and fractal dimension but have no clue how to implement them in deciding the sliding window size. Can someone provide me some guidelines or suggestions on any GitHub repo or package I can use to aid in building this?