97 Fails
97 - Fractal Analytics
Goal
Goal is to find the severals items demands using their historical demands. For some items, they share some common category, sub-categorical info. Tricky part is not all items have simmilar historical data, some have 9 months data and while others have only few weeks of data and we have predict 60 days into future.
Issues I had
-
Lack of proper testing framework or model.
-
Lack of knowledge on Time-Series Models. Example ARIMA.
-
Lack of experience of time series modelling.
Learning Sources
Time Series Analysis
-
Econometrics - Ani Katchova: Ani, convers the most part of theoritical Knowledge to understand and have a quick start in Time-Series Analysis. No Code - Just Theory.
-
Econometrics Academy: A bit abractly conveys understanding of AR, MA, ARMA and ARIMA, a bit deeper understanding. No Py Code - Just Theory.
-
PyData Conf - Jeffrey Yau: Theory + Hands on Code. Lots of Python Notebooks also available in Github.
-
PyCon 2017 - Aileen Nielsen: Complete Set. Theory + Basics + little Math + Code + Nice Notebooks in Github
Python Pkgs
-
PyFlux Documentation: Simple tools immediately start working with Time Series Analysis.
-
statsmodel - tsa (time series analysis)
Notebooks
- GitHub CookBook Resources